JP2004118856A - Information retrieval method and information retrieval system using agent - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To autonomously retrieve information, based on a simple direction from a user, and to easily provide the information out of a wide range of information source without requiring time and labor. <P>SOLUTION: In this information retrieving method using an agent, a plurality of agents set with an information retrieving condition inherent individually is exhibited (102), the agent is selected out of the plurality of agents, based on the information retrieving condition desired by the user (103), and information retrieval (111) desired by the user is carried out based thereon. <P>COPYRIGHT: (C)2004,JPO

Description

The present invention relates to an information search method and an information search device using an agent that searches for information of a large-scale storage device in the device or information of a large-scale storage device in a remote place by communication means.

(2) An information search device using an agent is described in, for example, Patent Document 1. FIG. 30 shows a block diagram of a conventional information retrieval apparatus using an agent.

The information retrieval apparatus using this conventional agent includes a voice input unit 3001, a unit 3003 for recognizing the input voice, a reliability determination unit 3004 for the voice recognition result, and a plurality of character screen data having different expressions. Data storage means 3007, a screen output means 3013, and a CPU 3006 for controlling these means, and reads character screen data according to the reliability level of the speech recognition result from the data storage means 3007 under the control of the CPU 3006. , And output from the screen output means.
JP-A-9-292895

However, the conventional information search described above has a problem that the user must set detailed information search conditions every time the information search is performed, which takes time and effort.

The present invention has been made to solve the above-mentioned conventional problems, and performs an information search autonomously based on a simple instruction of a user, and can easily retrieve information from a wide range of information sources without time and effort. An object of the present invention is to provide an information search method and an information search device using an agent capable of obtaining information.

The information search method using the agent of the present invention presents a plurality of predetermined agents in which individual information search conditions are individually set, and the agent determines an information search condition desired by a user from among the plurality of agents. The selected information is searched for by the user.

This method allows the user to obtain desired information without taking much time and effort.

(4) The specific information search condition is sequentially presented by voice or in the form of a list on a screen according to a voice from a user or an instruction from a remote controller.

(4) By this method, the user can confirm that the selected agent definitely has the information search condition desired by the user.

Further, when performing additional registration of the information search condition desired by the user, the information search condition desired by the user is incorporated into the existing information search condition, and an attempt is made to add the information search condition to the existing information search condition. When there is the same one as above, the additional registration is not performed.

This method allows the user to reliably search for the information he / she wants to search.

In addition, the information search condition additionally registered by the user is inspected to determine whether it is semantically inconsistent with the existing information search condition, and if the information search condition is inconsistent, the registration is rejected, or which is to be registered is determined. And

(4) According to this method, the user can easily determine that the selected agent is appropriate.

(4) When additionally registering information search conditions desired by the user, a voice dialogue between the user and the agent is used.

ユ ー ザ This method allows the user to reduce the burden of additional registration.

Further, when additionally registering the information search condition desired by the user, the information search condition desired by the user is input as a natural sentence, and the sentence is analyzed to extract conditions necessary for the information search, and the unique It was decided to add it to the table of information search conditions.

This method saves the user the time and effort of thinking about keywords.

(4) The agent that interacts with the user and performs a search in accordance with the information search condition is different from the agent that reads out information by the information search.

This method can keep the user interested in the agent.

{Circle around (4)} When the sentence of the information obtained by the information search matches a sentence existing in the agent in advance, an impression corresponding to the matching content is expressed to the user.

This method can increase the user's expectation of the search result.

(4) Further, the utterance content and utterance strategy of the dialogue with the user are determined according to the current emotional state of the user estimated from the history of the most recent dialogue with the user.

対 話 This method can generate a dialog that exactly matches the user's request.

内容 Further, the utterance content and the utterance strategy at the time of the dialogue with the user are determined according to the emotional state of the user estimated from the voice state of the user.

対 話 This method can generate a dialog that exactly matches the user's request.

内容 Further, the utterance content and utterance strategy at the time of the dialogue with the user are determined according to the emotional state of the user estimated from the expression of the user.

対 話 This method can generate a dialog that exactly matches the user's request.

In addition, the utterance content and the utterance strategy at the time of the dialogue with the user are determined according to the dialogue history with the user and the emotional state of the user that is comprehensively estimated from the voice state and expression of the user. And

対 話 This method can generate a dialog that exactly matches the user's request.

Also, a plurality of predetermined information sources are always searched using the information search conditions of the agent, and the search result is reported to the user at regular intervals.

This method saves the user from calling the agent.

Further, a plurality of predetermined information sources are always searched using the information search condition of the agent, and when the search result reaches a preset number, the result of the search is reported to the user. .

This method saves the user from calling the agent.

The information search condition of the agent is always used to search for a plurality of predetermined information sources. When the total amount of data obtained as a result of the search reaches a preset limit, the user is asked to perform the search. We decided to report the result of.

This method saves the user from calling the agent.

Further, when storing the information obtained by the information search in a storage area, the similarity to past information already stored is determined, and when similar information exists, the existence of the similar information is reported to the user. It was decided to.

This method can increase the user's affinity for the agent.

Further, when storing the information obtained by the information search in a storage area, the obtained information and the past information are analyzed and an abstracted one is generated as a set of mutually related keywords. It is used for similarity judgment, and the past information is deleted after a certain period of time.

This method can save the memory of the search result holding area and can compare the past and present search results at high speed.

Further, when storing the information obtained by the information search in a storage area, the obtained information and the past information are analyzed, and a set of broader terms of the extracted keywords is generated to determine similarity. Used, and the past information is deleted after a certain period of time.

(4) By this method, the memory of the search result holding area can be saved, and the similarity can be automatically determined irrespective of a slight difference in the keyword.

When storing the information obtained by the information search in a storage area, a set of keywords of both the new search result and the accumulated past search result is searched to find a pair of a verb and an object, Whether or not the pair of the verb and the object match each other is used as a criterion of similarity.

一致 This method can speed up the match determination.

Further, when new information is obtained by the information search, the newly searched information is analyzed to extract a theme which is the focus of the information, and is used to determine similarity with accumulated past information. I decided that.

This method enables automatic determination of similarity regardless of slight differences in keywords.

Also, the information search apparatus using the agent of the present invention includes a means for holding a plurality of predetermined agents, a means for storing information search conditions unique to each of the plurality of agents, and displaying the agent on a screen. And a means for selecting the displayed agent.

With this configuration, the user can obtain desired information without taking time and effort.

A means for interpreting an instruction from the user input through a microphone or a remote control receiver; a means for reading out an information search condition specific to the agent by a recorded voice or a synthesized voice in accordance with a result of the interpretation; , A means for displaying a list of information retrieval conditions unique to the device on a screen, and means for selectively executing the reading means and the means for displaying the list or simultaneously executing both of them.

With this configuration, the user can confirm that the selected agent definitely has the same interest as himself.

A means for extracting the information search condition from an input from a microphone or a remote control receiver; and determining whether the information search condition stored in the storage means and the input information search condition are identical. Means for generating a sentence that reports that the same thing has been removed, excluding the same thing when the same thing exists, means for outputting the sentence as synthesized speech, On the screen.

(4) With this configuration, the user can surely search for information to be searched.

The user further understands the meaning of the information search condition to be additionally registered, and the agent has already input the meaning of the information search condition already stored in the storage unit to the agent and newly input for the additional registration. Means for determining inconsistency with the meaning of the information search condition, means for rejecting additional registration from the user when the inconsistency is detected, and reporting the rejection of the additional registration to the user, or information for additional registration The apparatus includes means for inquiring of the user as to which of the search condition and the stored information search condition is to be selected, and re-registering the information in the information search condition storage area.

(4) With this configuration, the user can easily determine that the selected agent is appropriate.

A means for understanding the meaning of the utterance of the user using speech recognition means and a language analysis means including a language dictionary; and a means for determining whether or not the utterance of the user includes an information search condition desired by the user. And means for generating a dialogue strategy based on slot filling for extracting information search conditions from the user, and means for generating a synthesized speech in accordance with the dialogue strategy.

With this configuration, the user can reduce the burden of additional registration.

A means for inputting a sentence including an information search condition desired by the user; a means for comparing the sentence with a typical sentence template for analysis; In other words, the system includes means for performing analysis using a language dictionary and a language analysis algorithm, and means for extracting information search conditions desired by the user from the result of the analysis.

With this configuration, the user can save time and effort in thinking about keywords.

Further, the agent that interacts with the user and performs a search in accordance with the underlying information search condition, means for generating a strategy for selectively using the agent that reads out the information by the search, and means for operating the agent in accordance with the strategy The configuration is provided with.

With this configuration, the user's interest in the agent can be maintained.

A means for performing linguistic processing of the searched information; a means for understanding the meaning from the result of the linguistic processing; a means for expressing a theme of the information searched from the meaning as a network of keywords; Means for storing common sense inherent in the keyword network in the keyword network, means for determining mutual intimacy of the keyword network, means for generating a comment sentence corresponding to the result of the determination, And means for synthesizing.

With this configuration, the user's expectation of the search result can be increased.

A means for monitoring a conversation with the user and recording a conversation history; a means for estimating a current emotion of the user from the conversation history; and a means for generating a conversation strategy for guiding the emotion of the user to joy. Means for assembling the utterance according to the strategy, and changing the expression of the character when the character is displayed on the screen and the emotional expression of the synthesized voice.

With this configuration, it is possible to generate a dialogue that exactly matches the user's request.

A means for analyzing the prosody of the user's voice; a means for estimating the current emotion of the user from the prosody and the utterance content of the voice; and a means for generating a dialogue strategy for guiding the user's emotion to pleasure. Means for assembling an utterance in accordance with the dialogue strategy, and changing the expression of the character when the character is displayed on the screen and the expression of the emotion of the synthesized voice.

With this configuration, it is possible to generate a dialogue that exactly matches the user's request.

A means for recognizing the user's facial expression; a means for estimating the emotional state of the user from the facial expression; a means for generating a dialogue strategy that guides the user's emotion to pleasure; And a means for changing the expression of the character when the character is displayed on the screen.

With this configuration, it is possible to generate a dialogue that exactly matches the user's request.

A means for monitoring a dialogue with the user and recording a dialogue history; a means for analyzing a prosody of the user's voice; a means for recognizing the expression of the user; a dialogue history, the prosody and the utterance Means for estimating the current emotion of the user comprehensively from the combination of the content and the expression of the user, means for generating an interaction strategy that guides the emotion of the user to joy, and assembling an utterance according to the interaction strategy At the same time, the apparatus is provided with means for changing the expression of the character when the character is displayed on the screen and the emotional expression of the synthesized voice.

With this configuration, it is possible to generate a dialogue that exactly matches the user's request.

A means for searching while switching a plurality of information sources in accordance with information search conditions inherent in the agent; a means for storing information of a result of the search; Means for reporting the result of the search each time the time is reached.

(4) With this configuration, it is possible to save the user from calling the agent.

A means for searching while switching a plurality of information sources in accordance with an information search condition inherent in the agent; a means for storing information of a result of the search; and a method for storing the information when the stored number reaches a preset number. Means for reporting the result of the search.

(4) With this configuration, it is possible to save the user from calling the agent.

A means for performing a search while switching a plurality of information sources in accordance with an information search condition inherent in the agent; a means for storing information of a result of the search; a means for calculating the stored total amount of data; And a means for reporting the result of the search when the number reaches a preset upper limit.

(4) With this configuration, it is possible to save the user from calling the agent.

A means for storing information obtained by the information search; a means for judging similarity with past search result information stored in the storage means; and reporting the information on the judged similarity to the user. Means.

With this configuration, the user's affinity for the agent can be increased.

A means for storing the information obtained by the information search; a means for analyzing the information obtained by the search to abstract as a set of mutually related keywords; It is configured to include storage means for storing, means for causing an interruption each time a predetermined time is reached, and means for deleting the original text of the information which has been reported at the time of the interruption.

With this configuration, it is possible to save the memory of the search result holding area, and to compare the past and present search results at high speed.

A means for storing information obtained by the information search; a means for analyzing the information obtained by the search to abstract as a set of mutually associated keywords; and a means for determining a superordinate concept of the keyword. Storage means for storing both the superordinate concept and the original text of the search result; means for generating an interrupt each time a predetermined time is reached; and means for deleting the original text of the information that has been reported at the time of the interrupt. And a configuration including:

With this configuration, it is possible to automatically determine whether or not a match is made regardless of a slight difference between keywords.

Further, when information is obtained by the information search, means for storing a pair of a verb and an object in a set of both keywords in order to determine the similarity with the information of the accumulated past information search results. And a determining means for determining whether the pair of the verb and the object match.

With this configuration, similarity determination can be speeded up.

Further, when the information is obtained by the information search, means for analyzing the search result obtained by the information search to determine the similarity with the information of the accumulated past search results, and A means for extracting a theme of the information searched by using the information, a means for storing the theme together with the original text of the information as a search result, and a means for determining whether or not the theme matches are configured.

With this configuration, similarity can be automatically determined regardless of slight differences in keywords.

As described above, the present invention uses an agent that performs information retrieval autonomously based on a simple instruction of a user and can easily obtain applicable information from a wide range of information sources without taking time and effort. An information search method and an information search device can be provided.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First Embodiment)
FIG. 1 shows a flowchart of an information search method using an agent according to the first embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

First, in the information retrieval method using this agent, a list of images and characteristics of a plurality of agents acting on behalf of the user for information retrieval work is presented for the user (102), and the user obtains his / her own information from the images and characteristic descriptions. An agent who is likely to have a similar tendency to interest is selected (103). For example, in the case of "Father Agent," an image of an old man who seems to be stubborn and characteristics such as "Slightly stubborn. I am interested in politics and economy. We are weak for grandchildren and wives." If you want to make a search request for information such as, if you select this "father agent", it can be expected that almost the information that the user wants can be obtained.

Then the user selects that name or number. Only the selected agent remains and the other agents are deleted. As described above, each agent has a unique information search keyword (information search condition). For example, in the case of a businessman agent, words such as "salary," "business," and "selling" are set in advance as information search keywords. When the user requests to present an information retrieval keyword specific to the agent, a list of keywords is displayed on the display. In some cases, it is also effective to read out the registered contents using a synthetic voice or the like.

(4) When the agent is selected, the information search operation of the agent is started (111). If the user thinks that the presented information search keyword is not enough, the user can additionally register the information search keyword.

(4) At this time, it is checked whether the keyword overlaps or is inconsistent with the keyword registered in the past (107). If the same keyword is found in the past, the registration is stopped (108), and it is checked whether it is semantically inconsistent (109). If there is a contradiction, the one set by the user is registered with priority (110), Inconsistent past information search conditions are deleted.

Note that the information search condition does not need to be a word such as a keyword, but may be a sentence. When the set of keywords is completed, an information search operation is started (111), and the agent freely checks various channels of information sources (for example, Internet sites and text multiplex broadcasting) autonomously, and includes keywords. And present it to the user.

As described above, the information search method using an agent according to the first embodiment of the present invention has an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 2 shows a processing flow for inputting additional information search conditions by voice dialogue in the information search method using an agent according to the second embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

If the user thinks that the presented information search keyword (condition) is not sufficient, the user can request additional registration of the information search keyword by voice dialogue (201). Thereby, for example, the user can additionally register information search conditions even during driving.

{Circle around (2)} It is checked whether the keyword overlaps or is inconsistent with the keyword registered in the past (205). If the same keyword exists in the past, the registration is stopped (206), and it is checked whether the keyword is semantically inconsistent (207). If the keyword is inconsistent, the one set by the user is registered with priority (208). Inconsistent past information search conditions are deleted.

The information search condition does not need to be a word like a keyword, but may be a sentence. When the additional registration of the keyword (209) is completed, the agent freely checks various channels of information sources (for example, Internet sites and text multiplex broadcasting) autonomously, searches for the one containing the keyword, Present to the user.

As described above, the information search method using the agent according to the second embodiment of the present invention has an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 3 shows a processing flow for inputting an additional information search condition as a sentence in the information search method using an agent according to the third embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

If the user thinks that the presented information search keyword is not enough, the user can additionally register the information search keyword. First, the user inputs a sentence related to the information he or she wants to know from a remote controller or the like (303). For example, when "What did Prime Minister Moribuchi say yesterday?" Is input, the meaning of the input sentence is analyzed (304), the user really understands what he / she wants to do (305), and the information search keyword is input. Generate At this time, it is checked whether the keyword overlaps or is inconsistent with the keyword registered in the past (309). If there is the same keyword in the past, the registration is stopped (310), and it is checked whether it is semantically inconsistent (311). If there is a contradiction, the one set by the user is preferentially registered (312). Inconsistent past information search conditions are deleted.

The information search condition does not need to be a word like a keyword, but may be a sentence. After the additional registration of the keyword is completed (313), the agent freely checks various channels of the information source (for example, an Internet site or teletext), searches for a channel including the keyword, Present to the user.

As described above, the information search method using the agent according to the third embodiment of the present invention has an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 4 shows a processing flow in which an agent that reads out the searched information and an agent that interacts are separated in the information search method using an agent according to the fourth embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

{Circle around (1)} First, an agent automatically appears when a user calls (401) or a constraint condition such as the memory capacity of the user is about to be exceeded (403) is satisfied. When the agent is presented from the beginning, the agent starts to move vigorously under the above conditions. The agent autonomously checks various channels of the information source (for example, Internet sites and teletext) and searches for the one that contains the keyword. Before presenting to the user, the meaning of the search result is determined. Understand (404) and find the words that are in focus.

見 つ け Find the meaning of this focused word in the language dictionary on the memory of the agent, and generate a sentence expressing the corresponding impression. Since the sentence expressing this impression can be set in advance, the sentence does not necessarily need to be text-to-speech synthesis, but may be a recorded / reproduced speech. For example, when the focus of “exit” is extracted, the high school agent makes an impression corresponding to the meaning “Hey, just listen to this. Terrible!” (405). Thereafter, the agent switches to an agent specializing in reading (406). The read-only agent reads out the retrieved information using text-to-speech synthesis. Such alternation is continued until there is no more information retrieved.

As described above, the information search method using the agent according to the fourth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 5 shows a process flow of selecting an interaction strategy based on estimation of a user's emotion for smoothing interaction with the user in the information retrieval method using an agent according to the fifth embodiment of the present invention.

The operation of the information retrieval method using this agent will be described with reference to FIG.

First, after a one-turn interaction with the user (503), the success or failure of the interaction is recorded (504), a pattern of the interaction result is extracted (505), and the state of the user is estimated from the interaction history (506). I do.

For example, based on the record (504) of whether the conversations of the past several turns have succeeded or failed, it can be easily estimated (506) whether the user is "happy" or "angry". This is because "satisfaction" is "joy" and "slow" is "anger". When the user is "happy", the dialogue is advanced efficiently with a simple expression, but when the user is "angry", it is switched to a polite expression, the expression of the agent is changed to something that seems to be apologetic, and the estimated user Of the utterance method and the conversation strategy according to the emotion of the user (507).

As described above, the information search method using the agent according to the fifth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 6 shows a process flow of selecting a dialog strategy based on estimation of a user's emotion to facilitate a dialog with the user in the information search method using an agent according to the sixth embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

First, after a one-turn dialogue with the user (603), a situation analysis of the user's voice is performed (604), the estimated emotion of the user is extracted (605), and the current emotional state of the user is estimated from the voice quality of the user. (606).

For example, if the average pitch frequency is slightly higher and the duration of the ending is longer, the user is often happily. Conversely, if the average pitch frequency is lower, the overall power is increased, and the duration of the ending is shorter, Often "angry". Such estimation is performed for each conversation turn, and the estimation results are accumulated and used for emotion estimation. When the user is "happy", the dialogue is advanced efficiently with a simple expression, but when the user is "angry", it is switched to a polite expression, the expression of the agent is changed to something that seems to be apologetic, and the estimated user Of the utterance method and the dialog strategy in accordance with the emotion of the user (607).

As described above, the information search method using the agent according to the sixth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 7 shows a process flow of selecting a dialog strategy based on a user's expression to facilitate a dialog with a user in the information search method using an agent according to the seventh embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

First, after a one-turn interaction with the user (703), the user's facial expression is analyzed (704), the estimated emotion of the user up to the present is extracted (705), and the emotional state of the user is estimated from the user's facial expression. (706).

For example, if the eyes are thin and both ends of the mouth are raised, "pleasure" is common, the interval between both eyebrows is short, and if the mouth is slightly open, "anger" is common. . Such estimation is performed for each conversation turn, and the estimation results are accumulated and used for emotion estimation. When the user is "happy", the dialogue is advanced efficiently with a simple expression, but when the user is "angry", it is switched to a polite expression, the expression of the agent is changed to something that seems to be apologetic, and the estimated user Utterance method and dialogue strategy are selected according to the emotion of the user (707).

As described above, the information search method using the agent according to the seventh embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 8 shows a process flow of selecting an interaction strategy based on a user's voice quality, facial expression, and interaction history in order to smoothly interact with the user in the information search method using an agent according to the eighth embodiment of the present invention. .

The operation of the information search method using this agent will be described with reference to FIG.

First, after a one-turn dialogue with the user (803), the success or failure of the dialogue, the voice and facial expression of the user are recorded (804), and the dialogue history, voice status, and facial expression up to the present are taken out (805). Then, the emotional state of the user is estimated from the user's voice quality, facial expression, and the result of the conversation (806).

For example, if the result of the dialogue is “success”, the voice quality is “soft”, and the expression is “neutral”, the user is often satisfied, and conversely, the result of the dialogue fails and the voice quality is “angry”. If the expression is "angry", it is often "angry". Such estimation is performed for each conversation turn, and the estimation results are accumulated and used for emotion estimation. When the user is "happy", the dialogue is advanced efficiently with a simple expression, but when the user is "angry", it is switched to a polite expression, the expression of the agent is changed to something that seems to be apologetic, and the estimated user Utterance method and dialogue strategy are selected according to the emotion of the user (807).

As described above, the information search method using the agent according to the eighth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 9 shows a processing flow of an autonomous information search in the information search method using an agent according to the ninth embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

{Circle around (1)} First, the agent automatically obtains information including its own information search keyword and automatically changes the information channel CH one after another (903). As a result, when the search of all information channels is completed (905), the maximum number of registered information, the maximum memory size for storing information, or the search end condition for a preset time is satisfied (906), the search operation is interrupted. Then, the search information up to that point is transmitted to the user in a synthesized voice (908). Needless to say, in response to the user's call such as "Hey, please tell me OO", the search information up to that point can be transmitted to the user in a synthetic voice.

As described above, the information search method using the agent according to the ninth embodiment of the present invention has an excellent effect in that the user can easily perform the information search operation.

FIG. 10 shows a processing flow for referring to past search results in the information search method using an agent according to the tenth embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

First, when the information is searched (1002), the agent retrieves the first search result (1003), compares it with the past search results (1004), and determines whether there is any information similar to this information (1005). Examine the search result storage unit.

(4) When similar information is found, the fact is notified to the user (1006), and the search result is read out (1007). For example, when the stock price news is searched, the similarity of the past similar information is notified that "the stock has fallen again three days ago." And the searched information is read out. It is more efficient not to read past information unless there is a request from the user. Such processing is performed in all searches.

As described above, the information search method using the agent according to the tenth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 11 shows a processing flow for referring to past search results in the information search method using an agent according to the eleventh embodiment of the present invention.

The operation of the information search method using this agent will be described with reference to FIG.

{Circle around (1)} First, when the agent searches for information (1102), the agent checks the past search result storage unit for any similar information (1106). The determination of the similarity is the genre of the hit search keyword, the noun and object pair of the search condition, the superordinate concept of the search keyword, or the focus of the searched information. The information search keyword that has hit can of course be used for similarity determination, but using more of these concepts increases the probability of finding similar information. When similar information is found, the fact is notified to the user (1107), and the search result is read out (1108).

For example, if you are using a sportsman agent and find information that includes "tennis" in the "sports" genre at this time, before introducing it, the keyword "baseball" must be added to the "sports" genre. Is included, "Speaking of which, Nomo was a perfect match yesterday." And then reported past similar information, and then read out the retrieved tennis information.

動 As a pair of a verb and an object of the information search condition, for example, when “teach a nearby ramen shop” is input as the information search condition, “a ramen shop” and “teach” are extracted as a pair. Searching the past accumulated information, for example, if there is information found under the information search condition of "Tell me a delicious ramen shop" last week, this is extracted because it has the same pair, and the agent is I read the information of the ramen shop that can be found by searching at the moment. By doing so, the user can recall the experience at the ramen shop last week and review the current action.

上位 The upper concept differs in that it has a genre and a hierarchy. For example, the superordinate concept of “Ramen restaurant” is “Chinese restaurant” and the superordinate concept of “Chinese restaurant” is “place to eat”. Therefore, in the example of the ramen shop above, if there was information found by searching for "curry shop" last week, if you search for the keyword "ramen shop" at this time, the superordinate concept is the same "eating place" Is found, and the agent says, "Well, I went to a curry store last week." By doing so, the user can recall the experience at the curry store last week and review the current action.

焦点 Focus can summarize the searched information and extract it as the most frequently occurring word. The agent looks for past accumulated information containing this word and reports it to the user if found. For example, suppose that a businessman agent analyzes information retrieved using “crude oil” as a keyword, and “price increase” is extracted as a focus. The agent looks for information that focuses on "price information" in the past accumulated information, and finds information on information on residential land prices, for example, and says, "Speaking so, house prices went up last week." After that, it reads out the information on the price increase of the crude oil searched at the present time. By doing so, the user can foresee inflation based on the rise in house prices last week and the rise in crude oil prices at the present time, and can profit by taking action in accordance with the inflation.

By the way, it is more efficient not to read out the past information itself unless there is a request from the user, and it is enough to simply give an outline.

行 う Perform such processing in all searches.

As described above, the information search method using the agent according to the eleventh embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 12 is a block diagram of an information search device using an agent according to a twelfth embodiment of the present invention.

As shown in FIG. 12, an information retrieval apparatus using this agent comprises a central processing unit 1201, an agent display means 1202 which generally displays a list of selectable agents, and a display means. An interface circuit 1203 between the bus 1202 and the bus, an input unit 1204 generally composed of a microphone or a remote controller, an agent selecting unit 1205, a search keyword adding unit 1206, and an output unit of an audio device generally including a speaker. 1207, an information search condition storage unit 1208, which is generally a memory device for agent-specific information search conditions, an information search unit 1209, and a system bus 1210.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the information retrieval device using this agent The present device displays a plurality of agent images acting on behalf of the user for the information retrieval work and their characteristics on the display device 1202, and the user obtains his or her interests from the images and the description of the characteristics. Select an agent that is likely to have a similar tendency. For example, in the case of "Father Agent", an image of an old man who seems to be stubborn and characteristics such as "Slightly stubborn. I am interested in politics and economy. We are weak for grandchildren and wives" are displayed, so stock prices and yen exchange rates If you want to make a search request for information such as, if you select this "father agent", it can be expected that almost the information that the user wants can be obtained.

Therefore, the user selects the name or number using the input means 1204. When the input unit 1204 is a voice, the input unit 1204 includes a voice recognition device. Only the selected agent remains by the agent selecting means 1205, and the other agents are deleted. The agent selection means 1205 stores the name or number of this agent. Each agent has a unique information search keyword in the information search condition storage unit 1208. For example, in the case of a businessman agent, words such as "salary," "business," and "selling" are stored in advance as information search keywords.

(4) When a user requests the agent to present a unique information search keyword to the agent using the input unit 1204, a list of keywords is displayed on the display device 1202. In some cases, it is also effective to read out the registered contents from the output unit 1207 using a synthetic voice or the like. If the user thinks that the presented information search keyword is not enough, the user can additionally register the information search keyword. At this time, the central processing unit 1201 checks whether the keyword has any overlap or inconsistency with the keyword registered in the past.

If the same keyword exists in the past, the registration is stopped. If the keyword is semantically inconsistent, the one set by the user is prioritized and the inconsistent past information search condition is deleted. The information search condition need not be a word such as a keyword, but may be a sentence. When the keyword is set, the agent autonomously checks various channels of the information source (for example, an Internet site or a teletext broadcast), searches for the one containing the keyword, and presents it to the user.

As described above, the information search apparatus and the information search apparatus using the agent according to the twelfth embodiment of the present invention can provide an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 13 is a block diagram of an information retrieval condition unique to an agent and a decision unit 1300 for an information retrieval condition to be newly added in the information retrieval device using an agent according to the thirteenth embodiment of the present invention.

As shown in FIG. 13, the determination unit 1300 includes a sequence control unit 1301, an input unit 1302, a comparison determination unit 1303, a thesaurus dictionary 1304, an information search condition storage unit 1305 unique to an agent, and a gate unit 1306. It is composed of

The operation of the information search condition specific to the agent and the information search condition determination unit 1300 to be newly added in the information search device using the agent configured as described above will be described.

If the user thinks that the information search condition (keyword or sentence) of the twelfth embodiment is not sufficient, he or she uses the input means 1302 to input a new information search condition or information search condition expressed by a sentence. I do. This keyword or sentence is not sent directly to the information retrieval condition storage means 1305 but is temporarily stored in the gate means 1306. When the sequence control means 1301 knows that a keyword or sentence has been input by the notification from the input means 1302, it activates the comparison determination means 1303, and compares the input with the information already stored in the information search condition storage means 1305. I do.

At that time, not only a simple comparison but also the input and the stored one are set as a set of related words in the thesaurus dictionary 1304, and it is determined whether the sets have the same element. As a result, if it is determined that they are not the same, a command is issued from the sequence control means 1301 to transfer the input stored in the gate means 1306 to the information search condition storage means 1305.

{Circle around (5)} The comparison determining means 1303 also determines the consistency between the input and the existing information search condition. In other words, the condition of "high price" and the condition of "low price" are meaningless if they exist at the same time. When such a contradiction is found, the comparing and judging means 1303 notifies that there is a contradiction in the output means 1307, and the output means 1307 has a contradiction in synthesized speech or the like. The condition notifies the user that it has been deleted.

As described above, the new information search condition input by the user is automatically compared with the existing information search condition, and after checking the consistency and redundancy, the new information search condition is overwritten and registered with the contradictory one. If there is no inconsistency, it is additionally registered.

As described above, the information search device and the information search device using the agent according to the thirteenth embodiment of the present invention can provide an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 14 is a block diagram showing a case where a voice conversation is performed as the information search condition adding means 1400 in the information search device using the agent according to the 14th embodiment of the present invention.

As shown in FIG. 14, the information search condition adding unit 1400 includes a sequence control unit 1401, a voice recognition unit 1402, a dialogue management unit 1403, a comparison determination unit 1404, a thesaurus dictionary 1405, and an information search condition storage unit. 1406 and gate means 1407.

The operation of the information search apparatus using the agent configured as described above in the case where a voice conversation is performed as the information search condition adding means 1400 will be described.

If the user thinks that the information search condition (keyword or sentence) of the twelfth embodiment is not enough, the information search condition newly expressed by the information search keyword or sentence is input. This keyword or sentence is not sent directly to the information search condition storage means 1406, but is temporarily stored in the gate means 1407.

際 At this time, as a phenomenon peculiar to voice recognition, there are cases where erroneous recognition or rephrasing is required. The dialogue management means 1403 looks at the result notification from the speech recognition means 1402 and, if the likelihood of the recognition result is not sufficient, uses the speech synthesis means 1408 to request the user to re-speak. When the sequence control means 1401 knows that a valid keyword or sentence has been recognized by the notification from the dialog management means 1403, it activates the comparison judgment means 1404, and the input and the information have already been stored in the information search condition storage means 1406. Compare things. At this time, not only a simple comparison, but also both the input and the stored ones are set as a set of related words in the thesaurus dictionary 1405, and it is determined whether the sets have the same element. As a result, if it is determined that they are not the same, a command is issued from the sequence device to transfer the input stored in the gate unit 1407 to the information search condition storage unit 1406.

{Circle around (4)} The comparison determining means 1404 also determines the consistency between the input and the existing information search condition. In other words, the condition of "high price" and the condition of "low price" are meaningless if they exist at the same time. When such a contradiction is found, the comparison / judgment unit 1404 notifies that there is a contradiction in the voice synthesizing unit 1408, and the voice synthesizing unit 1408 has a contradiction. Notify users that they have been deleted.

As described above, the new information search condition input by the user is automatically compared with the existing information search condition, and is registered only when there is no problem after checking consistency and redundancy.

As described above, the information search device and the information search device using the agent according to the fourteenth embodiment of the present invention have an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 15 is a block diagram showing a case where a sentence is input as the information search condition adding means 1500 in the information search device using the agent according to the fifteenth embodiment of the present invention.

As shown in FIG. 15, the information search condition adding means 1500 includes a sequence control means 1501, an input means 1502, a syntax analysis means 1503, a meaning understanding means 1504, a keyword extraction means 1505, and a comparison and judgment means 1506. , An information retrieval condition storage unit 1507, a gate unit 1508, a response sentence generation unit 1509, and an output unit 1510.

An operation when inputting a sentence as the information search condition adding means 1500 of the information search device using the agent configured as described above will be described.

If the user thinks that the information search condition (keyword or sentence) of the twelfth embodiment is not sufficient, he or she uses the input means 1502 to input a new information search condition or information search condition expressed by a sentence. I do. Generally, input is often made with a remote controller. This keyword or sentence is not directly sent to the information search condition storage means 1507, but is temporarily stored in the gate means 1508. At this time, as a phenomenon peculiar to sentence input, there is a case where re-input is necessary due to erroneous writing or the like. The keyword is extracted by the keyword extracting means 1505 from the input decomposed into units close to words by the syntax analyzing means 1503 and the meaning understanding means 1504.

At this time, the syntax analysis means 1503 and the meaning understanding means 1504 sequentially transfer scores expressing the validity of the analysis to the next keyword extraction means 1505. The sequence control unit 1501 looks at the result of the keyword extraction unit 1505 that receives them, and if there is an error, uses the response sentence generation unit 1509 to request the user to input again. When the sequence control unit 1501 knows that a valid keyword has been extracted by the notification from the keyword extraction unit 1505, it activates the comparison determination unit 1506, and executes the input and the one already stored in the information search condition storage unit 1507. Compare. At this time, not only a simple comparison but also both the input and the stored one are set as a set of related words in the thesaurus dictionary 1511 and it is determined whether or not the sets have the same element.

As a result, when it is determined that they are not the same, a command is issued from the sequence control means 1501 to transfer the input stored in the gate means 1508 to the information search condition storage means 1507. Further, the comparison determining means 1506 also determines the consistency between the input and the existing information search condition. In other words, the condition of "high price" and the condition of "low price" are meaningless if they exist at the same time. When such a contradiction is found, the comparing and judging means 1506 notifies that there is a contradiction in the output means 1510, and the output means 1510 has a contradiction. To the user.

As described above, the new information search condition input by the user is automatically compared with the existing information search condition, and is registered only when there is no problem after checking consistency and redundancy.

As described above, the information search device and the information search device using the agent according to the fifteenth embodiment of the present invention have an excellent effect in that the user can save time and effort in setting information search conditions.

FIG. 16 is a block diagram showing a case where an agent is switched in the information search device using an agent according to the sixteenth embodiment of the present invention.

In FIG. 16, an information search device using this agent includes a speech recognition unit 1601, a syntax analysis unit 1602, a semantic understanding unit 1603, a dialogue management unit 1604, an information search unit 1605, an information collection unit 1606, It comprises a reception information storage unit 1607, an agent switching unit 1608, a recorded voice DB 1609, a text / speech synthesis unit 1610, an image generation unit 1611, an output switching unit 1612, and a speaker 1613.

The operation of switching the agent in the information retrieval device using the agent configured as described above will be described.

{Circle around (1)} First, when a call from the user through the voice recognition unit 1601 or a constraint condition such as an increase in the memory capacity of the agent itself is satisfied, the agent automatically appears through the image generation unit 1611. When the agent is presented from the beginning, the agent starts to move vigorously under the above conditions. The agent autonomously checks various channels of an information source (for example, an Internet site or text multiplex broadcasting), and information search conditions (keywords) of the agent itself are included in the information search means 1605 by the information search keyword. Before searching for the user, the information search unit 1605 finds a keyword that is the focus of the searched information.

対 話 Based on this, the dialogue management means 1604 selects a recorded voice expressing the impression associated with the focused keyword from the recorded voice database 1609. Since the sentence expressing this impression can be determined in advance, the sentence does not necessarily need to be text-to-speech synthesis, but may be a recorded / reproduced speech. For example, if the focus of “exit” is extracted, a high school girl agent makes an impression such as “Hey, just listen to this. Terrible!” Thereafter, the agent is switched to an agent specialized in reading aloud by the agent switching means 1608. The read-only agent reads out the retrieved information using text-to-speech synthesis. Such replacement is continued until there is no more information retrieved by the information retrieval means 1605.

As described above, the information search device using the agent according to the sixteenth embodiment of the present invention can provide an excellent effect in that the interest of the user can be maintained.

FIG. 17 is a block diagram of an information retrieval device using an agent according to a seventeenth embodiment of the present invention.

In FIG. 17, an information search device using this agent includes a speech recognition unit 1701, a syntax analysis unit 1702, a semantic understanding unit 1703, a dialogue management unit 1704, an information search unit 1705, an information collection unit 1706, It comprises a reception information storage unit 1707, an agent switching unit 1708, a recorded voice database 1709, a text voice synthesizing unit 1710, an image generation unit 1711, an output switching unit 1712, a speaker 1713, and a common sense database 1714. I have.

The operation of the information retrieval device using the agent configured as described above will be described.

{Circle around (1)} First, when a call from the user through the voice recognition unit 1701 or a constraint condition such as the memory capacity of the agent itself is likely to be satisfied is satisfied, the agent automatically appears through the image generation unit 1711. When the agent is presented from the beginning, the agent starts to move vigorously under the above conditions. The agent autonomously checks various channels of an information source (for example, an Internet site or text multiplex broadcast), and the information search condition (keyword) of the agent itself is included in the information search means 1705 including the information search keyword. Prior to presentation to the user, the focused information is extracted from the searched information, so the searched information is decomposed into morphemes by the preprocessing of the text-to-speech synthesis unit 1710. Then, the focus is extracted by the parsing means 1702 and the meaning understanding means 1703. When there are a plurality of focuses, the meaning understanding means narrows down the focus to one using the common sense database 1714.

対 話 Based on this, the dialogue management means 1704 selects a recorded voice expressing the impression associated with the focused keyword from the recorded voice database 1709. Since the sentence expressing this impression can be determined in advance, the sentence does not necessarily need to be text-to-speech synthesis, but may be a recorded / reproduced speech. For example, if the focus of “exit” is extracted, a high school girl agent makes an impression such as “Hey, just listen to this. Terrible!” Thereafter, the agent is switched to an agent specializing in reading aloud by the agent switching means 1708. The read-only agent reads out the retrieved information using text-to-speech synthesis. Such alternation is continued until there is no more information retrieved by the information retrieval means 1705.

Thus, the information search device using the agent according to the seventeenth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 18 is a block diagram showing an information retrieval apparatus using an agent according to an eighteenth embodiment of the present invention.

In FIG. 18, the information retrieval device using this agent includes a speech recognition unit 1801, a syntax analysis unit 1802, a semantic understanding unit 1803, a dialog management unit 1804, a dialog history database 1805, an emotion estimation unit 1806, Information collecting means 1807, received information storing means 1808, information searching means 1809, agent switching means 1810, recorded voice database 1811, text voice synthesizing means 1812, output switching means 1813, speaker 1814, image generation And means 1815.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the result of a one-turn dialogue with the user, which is made up of the output of the text-to-speech synthesizing unit 1812 in response to the input from the voice recognition unit 1801 (or the reverse combination), is recorded in the dialogue history database 1805. The emotion estimation unit 1806 estimates the emotional state of the user. For example, whether the user is "happy" or "angry" can be easily estimated based on whether the conversations in the past several turns have succeeded or failed. This is because "satisfaction" is "joy" and "slow" is "anger". If the user is "happy", the dialog management means 1804 proceeds the dialog efficiently with a slightly inexact expression, but if the user is "angry", the expression is switched to a polite expression and the image generation means 1815 apologizes for the expression of the agent. Replace with something that looks good.

Thus, the information search device using the agent according to the eighteenth embodiment of the present invention can provide an excellent effect in that the user's interest can be maintained.

FIG. 19 is a block diagram of an information search device using an agent according to a nineteenth embodiment of the present invention.

In FIG. 19, an information retrieval device using this agent includes a speech recognition unit 1901, a syntax analysis unit 1902, a meaning understanding unit 1903, a dialogue management unit 1904, an emotion estimation unit 1905, an information collection unit 1906, It comprises a reception information storage unit 1907, an information search unit 1908, an agent switching unit 1909, a recorded voice database 1910, a text-to-speech synthesis unit 1911, an output switching unit 1912, a speaker 1913, and an image generation unit 1914. ing.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the emotion state of the user is estimated using the emotion estimation means 1905 in response to the input from the voice recognition means 1901. For example, if the average pitch frequency is slightly higher and the duration of the ending is longer, the user is often happily. Conversely, if the average pitch frequency is lower, the overall power is increased, and the duration of the ending is shorter, Often "angry". If the user is "happy", the dialog management means 1904 proceeds with the dialog efficiently with a slightly inexact expression, but if the user is "angry", the expression is switched to a polite expression, and the image generation means 1914 apologizes for the expression of the agent. Replace with something that looks good.

Thus, the information search device using the agent according to the nineteenth embodiment of the present invention has an excellent effect in that the user's interest can be maintained.

FIG. 20 is a block diagram of an information search device using an agent according to a twentieth embodiment of the present invention.

In FIG. 20, an information retrieval apparatus using this agent includes a speech recognition unit 2001, a syntax analysis unit 2002, a meaning understanding unit 2003, a dialogue management unit 2004, an image recognition unit 2005, an emotion estimation unit 2006, Information collecting means 2007, received information storing means 2008, information searching means 2009, agent switching means 2010, recorded voice database 2111, text voice synthesizing means 2012, output switching means 2013, speaker 2014, image generation Means 2015 and consist.

The operation of the information retrieval device using the agent configured as described above will be described.

{Circle around (1)} First, the emotional state of the user is estimated using the emotional estimation unit 2006 for the user's facial expression coming in through the image recognition unit 2005 simultaneously with the input from the voice recognition unit 2001. For example, if the eyes are thin and both ends of the mouth are raised, "pleasure" is common, the interval between both eyebrows is short, and if the mouth is slightly open, "anger" is common. . If the user is “happy”, the dialog management means 2004 will proceed the dialog efficiently with a slightly inexact expression, but if the user is “anger”, switch to a polite expression and use the image generation means 2015 to apologize for the expression of the agent. Replace with something that looks good.

As described above, the information search device using the agent according to the twentieth embodiment of the present invention has an excellent effect in that the user's interest can be maintained.

FIG. 21 is a block diagram showing an information retrieval apparatus using an agent according to a twenty-first embodiment of the present invention.

In FIG. 21, an information retrieval device using this agent includes a speech recognition unit 2101, a syntax analysis unit 2102, a meaning understanding unit 2103, a dialogue management unit 2104, an image recognition unit 2105, an emotion estimation unit 2106, Information collecting means 2107, received information storing means 2108, information searching means 2109, agent switching means 2110, recorded voice database 2111, text voice synthesizing means 2112, output switching means 2113, speaker 2114, image generation Means 2115.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the user's emotional state is estimated using the emotion estimation hand 2106 with respect to the user's facial expression coming in through the image recognition means 2105 simultaneously with the input from the voice recognition means 2101. For example, if the result of the dialogue is “success”, the voice quality is “soft”, and the expression is “neutral”, the user is often satisfied, and conversely, the result of the dialogue fails and the voice quality is “angry”. If the expression is "angry", it is often "angry". If the user is "happy", the dialog management means 2104 proceeds the dialog efficiently with a slightly inexact expression, but if the user is "angry", the expression is switched to a polite expression and the image generation means 2115 apologizes for the expression of the agent. Replace with something that looks good.

As described above, the information search device using the agent according to the twenty-first embodiment of the present invention has an excellent effect in that the user's interest can be maintained.

FIG. 22 is a block diagram of an information search device using an agent according to a twenty-second embodiment of the present invention.

In FIG. 22, an information search device using this agent includes a sequence control unit 2201, an input unit 2202, a comparison and determination unit 2203, a thesaurus dictionary 2204, an information search condition storage unit 2205, a gate unit 2206, and an output unit. It comprises means 2207, information search means 2208, time monitoring means 2209, text-to-speech synthesis means 2210, received information storage means 2211, information collection means 2212, and a plurality of information sources 2213.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent seeks information including its own information search keyword, replaces the plurality of information sources 2213 one after another by the information collection unit 2212, and searches the information stored in the reception information storage unit 2211 by the information search unit 2208. To go. This search operation is monitored by the time monitoring means 2209, and when the time exceeds the required search time previously set in the sequence control means 2201, the search operation is terminated, and the search result is sent to the user by the text-to-speech synthesis means 2210. Tell

As described above, the information search device using the agent according to the twenty-second embodiment of the present invention has an excellent effect in that the user can easily perform the information search operation.

FIG. 23 is a block diagram showing an information retrieval apparatus using an agent according to a twenty-third embodiment of the present invention.

In FIG. 23, an information search device using this agent includes a sequence control unit 2301, an input unit 2302, a comparison and determination unit 2303, a thesaurus dictionary 2304, an information search condition storage unit 2305, a gate unit 2306, and an output unit. It comprises means 2307, information search means 2308, search number investigation means 2309, text-to-speech synthesis means 2310, received information storage means 2311, information collection means 2312, and a plurality of information sources 2313.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent seeks information including the information search keyword possessed by the agent, replaces a plurality of information sources 2313 one after another by the information collection unit 2312, and searches the information stored in the reception information storage unit 2311 by the information search unit 2308. To go. This search operation is monitored by the search number investigating means 2309, and when the number exceeds the upper limit set in advance in the sequence control means 2301, the search operation is terminated, and the search result is transmitted to the user by the text-to-speech synthesis means 2310. Tell

As described above, the information search device using the agent according to the twenty-third embodiment of the present invention has an excellent effect in that the user can easily perform the information search operation.

FIG. 24 is a block diagram of an information retrieval device using an agent according to a twenty-fourth embodiment of the present invention.

In FIG. 24, an information search device using this agent includes a sequence control unit 2401, an input unit 2402, a comparison and judgment unit 2403, a thesaurus dictionary 2404, an information search condition storage unit 2405, a gate unit 2406, an output unit It comprises means 2407, information searching means 2408, memory size checking means 2409, text-to-speech synthesizing means 2410, received information storing means 2411, information collecting means 2412, and a plurality of information sources 2413.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent seeks information including its own information search keyword, replaces a plurality of information sources 2413 one after another by the information collection unit 2412, and searches the information stored in the reception information storage unit 2411 by the information search unit 2408. To go. This search work is monitored by the memory size checking means 2409, and when the data stored in the reception information storage means 2411 exceeds the upper limit size set in the sequence control means 2401 in advance, the search work is terminated, The text-to-speech synthesis means 2410 informs the user of the search result at that time.

As described above, the information search device using the agent according to the twenty-fourth embodiment of the present invention has an excellent effect in that the user can easily perform the information search operation.

FIG. 25 is a block diagram showing an information retrieval apparatus using an agent according to a twenty-fifth embodiment of the present invention.

In FIG. 25, an information search device using this agent includes a sequence control unit 2501, an input unit 2502, a comparison and judgment unit 2503, a thesaurus dictionary 2504, an information search condition storage unit 2505, a gate unit 2506, and an output unit. Means 2507, information search means 2508, search upper limit condition investigation means 2509, text-to-speech synthesis means 2510, reception information storage means 2511, information collection means 2512, a plurality of information sources 2513, and past reception information. It is composed of a storage area 2514 and an agent speech database 2515.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent uses the information collecting means 2512 to autonomously search for information that matches the information search conditions of the agent from a plurality of information sources 2513.If there is no information similar to this information, The storage area 2514 of the reception information is checked. When similar information is found, the optimum speech is selected from the dialogue speech database 2515 of the agent, the fact is reported to the user through the output unit 2507, and the search result is read out using the text-to-speech synthesis unit 2510. For example, when searching for stock price news, the similarity of the past similar information is reported as "the stock fell again three days ago," and then the searched information is read out. Such an operation is continued until the search upper limit condition checking means 2509 commands the stop.

Thus, the information search device using the agent according to the twenty-fifth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 26 is a block diagram of an information search device using an agent according to a twenty-sixth embodiment of the present invention.

In FIG. 26, an information search device using this agent includes a sequence control unit 2601, an input unit 2602, a comparison and determination unit 2603, a thesaurus dictionary 2604, an information search condition storage unit 2605, a gate unit 2606, and an output unit. Means 2607, information search means 2608, search upper limit condition checking means 2609, text-to-speech synthesis means 2610, reception information storage means 2611, information collection means 2612, plural information sources 2613, and past reception information It is composed of a keyword genre storage area 2614, a keyword genre assigning means 2615, and an agent speech database 2616.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent uses the information collecting means 2612 to autonomously search for information that matches the information search condition of the agent itself from a plurality of information sources 2613, and searches for information similar to this information. The genre of the search keyword of the received information is determined by the keyword genre assigning means 2615, and the keyword genre storage area 2614 of the past received information is examined. If it is determined by the comparing and judging means 2603 that they are similar, the optimum speech is selected from the dialogue speech database 2616 of the agent, the fact is reported to the user through the output means 2607, and the search result is read out using the text-to-speech synthesis means 2610. For example, when a stock price news is searched, the similarity of the past similar information is reported as "the stock fell again three days ago", and then the searched information is read out. Such an operation is continued until the search upper limit condition checking means 2609 commands the stop.

As described above, the information search device using the agent according to the twenty-sixth embodiment of the present invention can provide an excellent effect in that the interest of the user can be maintained.

FIG. 27 is a block diagram of an information search device using an agent according to a twenty-seventh embodiment of the present invention.

In FIG. 27, an information search device using this agent includes a sequence control unit 2701, an input unit 2702, a comparison and determination unit 2703, a thesaurus dictionary 2704, an information search condition storage unit 2705, a gate unit 2706, and an output unit. Means 2707, information search means 2708, search upper limit condition checking means 2709, text-to-speech synthesis means 2710, reception information storage means 2711, information collection means 2712, a plurality of information sources 2713, and information received in the past. The storage area 2714 of the retrieval condition verb and object, the information retrieval condition verb and object extraction means 2715, and the dialogue speech database 2716 of the agent are provided.
The operation of the information search device using the agent configured as described above will be described.

First, the agent autonomously searches the information collecting means 2712 from a plurality of information sources 2713 for information that matches the information search condition of the agent itself, and searches for information similar to this information. The extraction means 2715 extracts the verb and the object of the information retrieval condition thus obtained, and checks the storage area 2714 of the verb and the object of the information retrieval condition received in the past. If the comparison and judgment means 2703 judges that the speech is similar, the optimum speech is selected from the dialogue speech database 2716 of the agent, the fact is reported to the user through the output means 2707, and the search result is read out using the text speech synthesis means 2710. For example, when searching for stock price news, the similarity of the past similar information is reported as "the stock fell again three days ago," and then the searched information is read out. Such an operation is continued until the search upper limit condition checking unit 2709 orders a stop.

As described above, the information search device using the agent according to the twenty-seventh embodiment of the present invention can provide an excellent effect in that the interest of the user can be maintained.

FIG. 28 is a block diagram of an information retrieval device using an agent according to a twenty-eighth embodiment of the present invention.

In FIG. 28, an information search device using this agent includes a sequence control unit 2801, an input unit 2802, a comparison and judgment unit 2803, a thesaurus dictionary 2804, an information search condition storage unit 2805, a gate unit 2806, and an output unit. Means 2807, information search means 2808, search upper limit condition checking means 2809, text-to-speech synthesis means 2810, reception information storage means 2811, information collection means 2812, a plurality of information sources 2813, It is composed of a search keyword super concept storage area 2814, a search keyword super concept addition means 2815, and an agent speech voice database 2816.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent autonomously searches the information collecting means 2812 from a plurality of information sources 2813 for information that matches the information search condition of the agent itself, but searches for information similar to this information. The higher-level concept of the search keyword of the received information is assigned by the assigning means 2815, and the upper-level concept storage area 2814 of the search keyword of the past received information is checked.

(4) When the comparison / judgment means 2803 judges that the speech is similar, the optimum speech is selected from the dialogue speech database 2816 of the agent, the fact is reported to the user through the output means 2807, and the search result is read out using the text speech synthesis means 2810. The superordinate concept differs in having a genre and a hierarchy. For example, the superordinate concept of “Ramen restaurant” is “Chinese restaurant” and the superordinate concept of “Chinese restaurant” is “place to eat”.

Therefore, in the above example of a ramen shop, if there is information found by searching for “curry shop” last week, if you search with the keyword “ramen shop” at this time, the superordinate concept is the same “eating place” Is found, and the agent says, "Well, I went to a curry store last week." By doing so, the user can recall the experience at the curry store last week and review the current action. The present information search device continues such an operation until the search upper limit condition checking means 2809 commands the stop.

As described above, the information search device using the agent according to the twenty-eighth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

FIG. 29 is a block diagram of an information search device using an agent according to a twenty-ninth embodiment of the present invention.

In FIG. 29, an information search device using this agent includes a sequence control unit 2901, an input unit 2902, a comparison and judgment unit 2903, a thesaurus dictionary 2904, an information search condition storage unit 2905, a gate unit 2906, and an output unit. Means 2907, information search means 2908, search upper limit condition checking means 2909, text-to-speech synthesis means 2910, reception information storage means 2911, information collection means 2912, a plurality of information sources 2913, and past reception information. It is composed of a focus storage area 2914, a focus extraction means 2915 for search information, and an agent speech database 2916.

The operation of the information retrieval device using the agent configured as described above will be described.

First, the agent uses the information collecting means 2912 to autonomously search for information matching the information search condition of the agent itself from a plurality of information sources 2913, and searches for information similar to this information. The focus of the obtained information is estimated by the estimating means 2915, and the focus storage area 2914 of the past received information is examined. If it is determined by the comparing and judging means 2903 that they are similar, the optimum speech is selected from the dialogue speech database 2916 of the agent, the fact is reported to the user through the output means 2907, and the search result is read out using the text-to-speech synthesis means 2910.

For example, suppose that a business agent analyzes information retrieved using "crude oil" as a keyword, and "price rise" is extracted as a focus. The agent searches for information that focuses on "price information" in the past accumulated information, and finds information on housing land price information, for example, and says, "Speaking of which, the house price rose last week." Then, it reads out the information on the price increase of the crude oil searched at the present time.

By doing so, the user can foresee inflation from the rise in house prices last week and the current rise in crude oil prices, and benefit from taking action accordingly. The present information search device continues such an operation until the search upper limit condition checking means 2909 commands the stop.

Thus, the information search device using the agent according to the twenty-ninth embodiment of the present invention has an excellent effect in that the interest of the user can be maintained.

An information search method and an information search device using an agent according to the present invention perform an information search autonomously based on a simple instruction of a user, and can easily obtain relevant information from a wide range of information sources without time and effort. And is useful for an information search method and an information search apparatus for performing information search using an agent.

FIG. 4 is a flowchart of an information search method using an agent according to the first embodiment of the present invention. Processing flow diagram for inputting additional information search conditions by voice dialog in the information search method using an agent according to the second embodiment of the present invention Processing flow diagram for inputting an additional information search condition as a sentence in the information search method using an agent according to the third embodiment of the present invention In the information search method using an agent according to the fourth embodiment of the present invention, a processing flow diagram in which an agent that reads out the searched information and an agent that interacts are separated. In the information retrieval method using an agent according to the fifth embodiment of the present invention, a processing flow chart of selection of a dialog strategy based on estimation of a user's emotion for smoothing a dialog with the user In the information retrieval method using an agent according to the sixth embodiment of the present invention, a processing flow chart of selecting a dialog strategy based on estimation of a user's emotion to facilitate a dialog with the user In the information search method using an agent according to the seventh embodiment of the present invention, a processing flow diagram of selecting a dialog strategy based on a user's expression for smoothing a dialog with a user. In the information retrieval method using an agent according to the eighth embodiment of the present invention, a processing flow diagram of selecting a dialog strategy based on a user's voice quality, facial expression and a dialog history to facilitate a dialog with the user. In the information search method using an agent according to the ninth embodiment of the present invention, a processing flowchart of autonomous information search In the information search method using an agent according to the tenth embodiment of the present invention, a processing flowchart for referring to past search results In the information search method using an agent according to the eleventh embodiment of the present invention, a processing flowchart for referring to past search results Block diagram of an information retrieval apparatus using an agent according to a twelfth embodiment of the present invention FIG. 16 is a block diagram of a determination unit of an information search condition specific to an agent and a newly added information search condition in an information search device using an agent according to a thirteenth embodiment of the present invention; FIG. 14 is a block diagram of a case where a voice dialogue is performed as an information search condition adding unit in the information search device using an agent according to the 14th embodiment of the present invention; FIG. 15 is a block diagram of a case where a sentence is input as an information search condition adding unit in the information search device using an agent according to the fifteenth embodiment of the present invention; Block diagram of an agent switching unit in an information retrieval device using an agent according to a sixteenth embodiment of the present invention Block diagram of an information retrieval device using an agent according to a seventeenth embodiment of the present invention Block diagram of an information retrieval device using an agent according to an eighteenth embodiment of the present invention Block diagram of an information search device using an agent according to a nineteenth embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twentieth embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twenty-first embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twenty-second embodiment of the present invention Block diagram of an information search device using an agent according to a twenty-third embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twenty-fourth embodiment of the present invention Block diagram of an information search device using an agent according to a twenty-fifth embodiment of the present invention Block diagram of an information search device using an agent according to a twenty-sixth embodiment of the present invention Block diagram of an information search device using an agent according to a twenty-seventh embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twenty-eighth embodiment of the present invention Block diagram of an information retrieval device using an agent according to a twenty-ninth embodiment of the present invention Block diagram of an information retrieval device using a conventional agent

Explanation of reference numerals

1201 Central processing unit 1202 Agent display means 1203 Interface circuit,
1204, 1302, 1502, 2202, 2302, 2402, 2502, 2602, 2702, 2802, 2902 Input means 1205 Agent selection means 1206 Search keyword addition means 1207, 1510, 2207, 2307, 2407, 2507, 2607, 2707, 2807, 2807, 2907 Output means 1208, 1305, 1406, 1507, 2205, 2305, 2405, 2505, 2605, 2705, 2805, 2905 Information search condition storage means 1209, 1605, 1705, 1809, 1908, 2009, 2109, 2208, 2308, 2408 , 2508, 2608, 2708, 2808, 2908 Information search means 1301, 1401, 1501, 2201, 2301, 2401, 2501, 2601, 2701, 2801, 2901 Sequence control means 1303, 1404, 1506, 2203, 2303, 2403, 2503 , 2603, 2703, 2803, 2903 Comparison judgment means 1304, 1405, 2204, 2304, 2504, 2604, 2704, 2804, 2904 Thesaurus dictionaries 1402, 1601, 1701, 1801, 1901, 2001, 2101 Speech recognition means 1403, 1604, 1704, 1804, 1904, 2004, 2104 Dialogue management means 1503, 1602, 1702, 1802, 1902, 2002, 2102 Parsing means 1504, 1603, 1703, 1803, 1903, 1903, 2003, 2 103 Meaning understanding means 1505 Keyword extraction means 1509 Response sentence generation means 1606, 1706, 1807, 1906, 2007, 2107, 2212, 2312, 2412, 2512, 2612, 2712, 2812, 2811, 2912 Information collection means 1607, 1707, 1808 , 1907, 2008, 2108, 2211, 2311, 2411, 2511, 2611, 2711, 2811, 2911 Received information storage means 1608, 1708, 1810, 1909, 2010, 2110 Agent switching means 1609 Recording voice DB
1610, 1710, 1911, 2210, 2310, 2410, 2510, 2610, 2710, 2810, 2910 Text-to-speech synthesis means 1611, 1711, 1815, 1914, 2015, 2115 Image generation means 1709, 1811, 1910, 2011, 2111 Recorded voice Database 1714 Common sense database 1805 Dialogue history database 1806, 1905, 2006, 2106 Emotion estimation means 1812, 1911, 2012, 2112, 2210, 2310 Text-to-speech synthesis means 2005, 2105 Image recognition means 2209 Time monitoring means 2313, 2213, 2413, 2513 , 2613, 2713, 2813, 2913 Multiple information sources 2309, 2509, 2609, 2709, 2809, 2909 Search upper limit search means 2514 Past received information storage area 2515, 2616, 2716, 2816, 2916 Agent speech database of agents 2614 Keyword genre storage area of past received information 2615 Keyword genre assigning means 2714 Storage area of verb and object of search condition of past received information 2715 Extraction means of verb and object of search condition 2814 Past reception Top-level concept storage area for search keywords in communication information 2815 Means for assigning top-level concepts for search keywords 2914 Focus storage area for past received information 2915 Focus extraction means for search information

Claims (38)

Presenting a plurality of predetermined agents each having a unique information search condition, selecting the agent from the plurality of agents according to an information search condition desired by a user, and performing an information search desired by the user; An information retrieval method using an agent characterized by the following. When the additional registration of the information search condition desired by the user is performed, the information search condition desired by the user is incorporated into the existing information search condition, and is the same as that to be added to the existing information search condition. 2. The information retrieval method using an agent according to claim 1, wherein the additional registration is not performed when there is any of the following. Checking whether the information search condition additionally registered by the user is semantically inconsistent with the existing information search condition, and rejecting the registration if it is inconsistent, or determining which is to be registered; An information retrieval method using an agent according to claim 2, wherein 3. The information search method using an agent according to claim 2, wherein when the information search condition desired by the user is additionally registered, a voice dialogue between the user and the agent is used. When additionally registering the information search condition desired by the user, the information search condition desired by the user is input as a natural sentence, and the sentence is analyzed to extract a condition necessary for the information search, and the unique information search is performed. 2. The information search method using an agent according to claim 1, wherein the information is added to a condition table. 2. The information search using an agent according to claim 1, wherein the agent that interacts with the user and performs a search in accordance with an information search condition is different from an agent that reads out information by the information search. Method. 2. The agent according to claim 1, wherein when a sentence of the information obtained by the information search matches a sentence existing in the agent in advance, an impression corresponding to the matched content is expressed to the user. Information retrieval method using. The agent according to claim 1, wherein an utterance content and an utterance strategy of the dialogue with the user are determined according to a current emotional state of the user estimated from a history of the most recent dialogue with the user. The information retrieval method used. The agent according to claim 1, wherein the utterance content and the utterance strategy at the time of the dialogue with the user are determined according to the emotional state of the user estimated from the voice state of the user. Information retrieval method. The information retrieval using the agent according to claim 1, wherein the utterance content and the utterance strategy at the time of the dialogue with the user are determined according to the emotional state of the user estimated from the expression of the user. Method. The utterance content and the utterance strategy at the time of the dialogue with the user are determined according to the dialogue history with the user and the emotional state of the user which is comprehensively estimated from the voice state and expression of the user. An information retrieval method using the agent according to claim 1. 2. The information using an agent according to claim 1, wherein a plurality of predetermined information sources are always searched using the information search condition of the agent, and a search result is reported to the user at regular time intervals. retrieval method. A plurality of predetermined information sources are always searched using the information search condition of the agent, and when the search result reaches a preset number, the search result is reported to the user. An information search method using the agent according to claim 1. A plurality of predetermined information sources are always searched using the information search condition of the agent, and when the total amount of data obtained as a result of the search reaches a preset limited capacity, the result of the search is given to the user. 3. The information search method using an agent according to claim 2, wherein the information is reported. When storing the information obtained by the information search in a storage area, determining similarity with past information already stored, and when similar information exists, reporting the existence of the similar information to the user. The information retrieval method using an agent according to claim 1, wherein: When the information obtained by the information search is stored in a storage area, the obtained information and the past information are analyzed, and an abstracted one is generated as a set of mutually related keywords to generate similarity. 2. The information search method using an agent according to claim 1, wherein the information is used for determination, and the past information is deleted after a lapse of a predetermined time. When storing the information obtained by the information search in a storage area, the obtained information and the past information are analyzed, and a set of broader terms of the extracted keywords is generated and used for similarity determination. 2. The method according to claim 1, wherein the past information is deleted after a predetermined time has elapsed. When storing information obtained by the information search in a storage area, a set of a verb and an object is found by examining a set of keywords of both the new search result and the accumulated past search result, and a pair of a verb and an object is found. 2. The information retrieval method using an agent according to claim 1, wherein whether or not the pair of the verb and the object is identical is used as a criterion of similarity. When new information is obtained by the information search, analyzing the newly searched information, extracting a theme that is the focus of the information, and using the theme to determine similarity with accumulated past information. An information retrieval method using an agent according to claim 1, wherein Means for holding a plurality of predetermined agents, means for storing information search conditions unique to each of the plurality of agents, means for displaying the agents on a screen, and means for selecting the displayed agents An information retrieval device using an agent, comprising: Means for extracting the information search condition from an input from a microphone or a remote control receiver, and means for determining whether the information search condition stored in the storage means and the input information search condition are the same. Means for generating a sentence that reports that the same thing has been removed, excluding the same thing when the same thing exists, means for outputting the sentence as synthesized speech, and displaying the sentence on a screen 21. An information retrieval apparatus using an agent according to claim 20, further comprising: means for displaying the information. A means for understanding the meaning of the information search condition that the user intends to additionally register, the meaning of the information search condition already stored in the storage means in the agent, and the information search newly input for additional registration Means for determining inconsistency with the meaning of the condition, means for rejecting the additional registration from the user when the inconsistency is detected, and reporting the rejection of the additional registration to the user, or information search condition to be additionally registered 21. An agent according to claim 20, further comprising: means for inquiring of said user which of said stored information search conditions is to be selected and re-registering said information search conditions in a storage area of said information search conditions. Information retrieval device. Means for understanding the meaning of the utterance of the user using speech recognition means and language analysis means including a language dictionary, and means for determining whether the information search condition desired by the user is included in the utterance of the user, 21. The agent according to claim 20, further comprising: means for generating a dialogue strategy based on slot filling for extracting information search conditions from the user; and means for generating a synthesized speech according to the dialogue strategy. Information retrieval device used. A means for inputting a sentence including an information search condition desired by the user; a means for comparing the sentence with a typical sentence template for analysis; and an input which does not match the typical sentence template. 21. The agent according to claim 20, further comprising: means for performing analysis using a language dictionary and a language analysis algorithm; and means for extracting information search conditions desired by the user from a result of the analysis. Information retrieval device. The agent that interacts with the user and performs a search in accordance with an underlying information search condition; a unit that generates a strategy for selectively using the agent that reads out information by the search; and a unit that operates the agent in accordance with the strategy. 21. An information retrieval apparatus using an agent according to claim 20, wherein: Means for performing linguistic processing of the searched information; means for understanding the meaning from the result of the linguistic processing; means for expressing the theme of the information searched for from the meaning as a network of keywords; Means for storing the common sense of the keyword in the keyword network, means for determining mutual intimacy of the keyword network, means for generating an impression sentence corresponding to the result of the determination, and synthesizing the impression sentence 21. An information retrieval apparatus using an agent according to claim 20, further comprising means for performing an operation. Means for monitoring the interaction with the user and recording an interaction history, means for estimating the current user's emotion from the interaction history, and means for generating an interaction strategy for guiding the user's emotion to joy, 21. The agent according to claim 20, further comprising means for assembling the utterance according to the strategy, and means for changing the expression of the character when the character is displayed on the screen and the emotional expression of the synthesized voice. Information retrieval device. Means for analyzing the prosody of the user's voice, means for estimating the current emotion of the user from the prosody and the utterance content of the voice, and means for generating a dialogue strategy for guiding the user's emotion to pleasure, 21. The agent according to claim 20, further comprising means for assembling an utterance in accordance with the dialogue strategy, and means for changing the expression of the character when the character is displayed on a screen while expressing the emotion of the synthesized voice. Information retrieval device. Means for recognizing the user's facial expression, means for estimating the emotional state of the user from the facial expression, means for generating a dialogue strategy that guides the user's emotion to pleasure, and assembling and combining utterances according to the dialogue strategy. 21. The information retrieval apparatus using an agent according to claim 20, further comprising: means for changing a facial expression of the character when the character is displayed on a screen and a voice emotion expression. Means for monitoring the conversation with the user and recording a conversation history, means for analyzing the prosody of the user's voice, means for recognizing the expression of the user, the conversation history, the contents of the prosody and the utterance, Means for estimating the current user's emotions comprehensively from the combination of the user's facial expressions, and means for generating a dialogue strategy that guides the user's emotion to joy, and assembling an utterance according to the dialogue strategy, 21. The information retrieval apparatus using an agent according to claim 20, further comprising means for changing an expression of the emotion of the synthesized voice and a facial expression of the character when the character is displayed on the screen. Means for searching while switching a plurality of information sources in accordance with information search conditions inherent in the agent, means for storing information of the result of the search, each time a predetermined time elapses, or at a certain time 21. The information retrieval apparatus using an agent according to claim 20, further comprising: means for reporting a result of the retrieval each time the retrieval is performed. Means for searching while switching a plurality of information sources according to information search conditions inherent in the agent, means for storing information of the result of the search, and means for executing the search when the stored number reaches a preset number. 21. The information retrieval apparatus using an agent according to claim 20, further comprising means for reporting a result. Means for searching while switching a plurality of information sources according to the information search condition inherent in the agent, means for storing information of the result of the search, means for calculating the stored total amount of data, 21. The information search apparatus using an agent according to claim 20, further comprising: means for reporting a result of the search when a set upper limit is reached. Means for storing information obtained by the information search, means for determining similarity to information of past search results stored in the storage means, and means for reporting the information determined for the similarity to the user. An information retrieval apparatus using an agent according to claim 20, comprising: Means for storing information obtained by the information search, means for analyzing the information obtained by the search and abstracting it as a set of mutually related keywords, and storing both the set of keywords and the original sentence 21. The apparatus according to claim 20, further comprising a storage unit, a unit that causes an interruption at a predetermined time, and a unit that deletes the original text of the information that has been reported at the time of the interruption. An information retrieval device using an agent. Means for storing information obtained by the information search, means for analyzing the information obtained by the search and abstracting it as a set of mutually related keywords, and means for determining a superordinate concept of the keyword, Storage means for storing both the superordinate concept and the original text of the search result, means for causing an interruption every predetermined time, and means for deleting the original text of the information that has been reported at the time of the interruption. An information retrieval apparatus using an agent according to claim 20, comprising: Means for storing a pair of a verb and an object in a set of both keywords in order to determine the similarity with the information of the accumulated past information search result when the information is obtained by the information search; 21. The information retrieval apparatus using an agent according to claim 20, further comprising: a determination unit configured to determine whether a pair of a verb and an object match. Means for analyzing the search result obtained by the information search to determine the similarity with the information of the accumulated past search result when the information is obtained by the information search; and using the result of the analysis. Means for extracting a theme of the searched information, means for storing the theme together with the original text of the information as a search result, and means for determining whether or not the theme matches. Item 20. An information retrieval device using the agent according to Item 20.

Images