JP2002287793A - Command processing device, command processing method, and command processing program - Google Patents

Abstract

(57) [Summary] [PROBLEMS] A command processing device and a command capable of accurately grasping a user's request, reducing a user's input burden, and enabling a conversation considering a psychological distance held by the user with respect to the device. To provide a processing method and a command processing program. SOLUTION: In response to a user's utterance request, a certainty factor that quantifies the likelihood of the intention of the user and a psychological distance sensation that quantifies the psychological distance that the user has in the command processing device. To change the content and expression of the response sentence. Further, when the certainty factor is high, the user is not confirmed before executing the command, but is notified after the command is executed, thereby reducing stress when the user operates the command processing device.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a command processing device, a command processing method, and a command processing program, for example, to processing a command obtained from a user by voice recognition.

[0002]

2. Description of the Related Art With the rapid development of speech recognition technology in recent years, electronic devices that operate by interpreting the contents of natural language (called command request utterances) input using words that a user uses everyday are used. It has become. For example, there are a television that adjusts the channel to 5 when the user speaks “I want to see 5 channels”, and a word processor that can input and edit a document by the user's voice.

For example, a command processing device that analyzes and interprets a command input by a user in a natural language by voice includes, for example, an IIF (Intelligent Interface).
ce) There is an analysis engine. FIG. 10 is a diagram showing an example of the configuration of a conventional IIF analysis engine. An input sentence 201 input by a user via a microphone installed on an operated electronic device such as a television is sent to a language processing analysis component 202. Language processing analysis component 2
A step 02 divides the input sentence into clauses and acquires sentence structure information of the input sentence.

[0004] The case matching processing component 203 compares the sentence structure information obtained from the language processing analysis component 202 with the case registered in the case dictionary 204, obtains execution command information 205 from the user's command request utterance, and outputs it. I do. The execution command information 205 causes the operated electronic device to execute a command, such as tuning a television channel, turning off a light, or underlining a document being created by a word processor.

[0005] As a command processing apparatus for interacting with a user, there is, for example, a "knowledge information processing system" disclosed in Japanese Patent Laid-Open Publication No. 1-258036. This knowledge information processing system performs a matching process on a user's input sentence, and displays the searched rule and its certainty factor. Then, the user can change the certainty factor, and by reflecting the changed certainty factor information, it is possible to obtain a more desired result in the next and subsequent inferences.

Another example is disclosed in Japanese Patent Application Laid-Open No. 9-23727.
There is an "in-depth dialogue system" for public relations No. 8. This system is a system for coping with a situation in which a dialog between a user and an agent (command processing device) does not engage with a scenario prepared in advance by replacing the scenario from the beginning and replacing the scenario.

[0007]

However, the conventional command processing apparatus based on natural language input requires the user to execute a command request with a high degree of certainty, which may be reported afterwards, before the execution. In some cases, it may be checked whether or not the command can be executed, and the user may be stressed excessively. For example, if you want the background color to be blue in a word processor, execute "" Background color blue "" in response to a user's explicit request for "Background color blue". The confirmation before executing the command was performed.

Further, in response to a vague user request with low confidence in the analysis result, no specific command proposal has been made through dialogue. For example, in response to a vague request from a user to “make it stand out,” simply displaying a list of dialogues and displaying candidates, and making specific proposals to the user as to which of the candidates should be specifically selected I didn't.

Furthermore, even if the user repeatedly uses the device in the same manner, the intention of the user cannot be understood unless the user makes a specific request to the device. For example, if the user requests "do as usual" or "is there", the device cannot operate, and the word processor should be "underlined" or the television should be I had to make a specific request, such as "set channel 1".

[0010] In addition, the user cannot determine the position in which the user wants to contact the device, and in any case, confirms the command with a fixed expression such as "Is it OK?" To do so, the dialogue was monotonous and mechanical.

SUMMARY OF THE INVENTION An object of the present invention is to provide a command processing device capable of accurately grasping a user's request, reducing the input load on the user, and enabling a dialog taking into account the psychological distance the user has with the device. , A command processing method, and a command processing program.

[0012]

According to the present invention, there is provided an input sentence obtaining means for obtaining an input sentence input by a user in a natural language, and an input sentence obtained by the input sentence obtaining means. Command information obtaining means for obtaining command information from the input sentence, and when the command information obtained by the command information obtaining means lacks information necessary to specify an execution command to be executed, Insufficient information inferring means for inferring missing information, and when the command information includes information necessary to identify the execution command, identify the command from the command information, the command information When the information necessary for specifying the execution command is insufficient, the execution command is used by using the command information and the information inferred by the insufficient information inference means. Execution command identification means for identifying the command, the degree of matching of the command specified by the execution command identification means to the command intended by the input sentence, the matching degree inference means, and the execution command identification means When creating a response sentence corresponding to the specified execution command, a response sentence creating unit that creates by changing the expression of the response sentence according to the degree of matching inferred by the matching degree inference unit; There is provided a command processing device comprising: a presentation unit for presenting a response sentence created by a response sentence creation unit (first configuration). Furthermore, in order to achieve the above object, the present invention provides an input sentence obtaining unit for obtaining an input sentence input in a natural language from a user, and executing the input sentence obtained from the input sentence obtaining unit. Command information acquisition means for acquiring command information used when specifying a command, and the input sentence acquired by the input sentence acquisition means, from which the user is holding the object operated by the execution command When creating a response sentence corresponding to the command specified by using the command information acquired by the command information acquired by the psychological distance inference means for inferring the degree of psychological distance feeling, the psychological distance Response sentence creating means for changing the expression of the response sentence according to the degree of psychological distance inferred by the feeling inference means, and a response sentence created by the response sentence creation means And presenting means for presenting, to provide a command processing apparatus characterized by comprising a (second configuration). Further, in order to achieve the above object, the present invention provides an input sentence acquiring means for acquiring an input sentence input by a user in a natural language, and command information from the input sentence acquired by the input sentence acquiring means. Command information acquisition means to be acquired, and in the case where information necessary for specifying an execution command to be executed is insufficient in the command information acquired by the command information acquisition means, infer the missing information. Shortage information inference means to perform, if the command information includes information necessary to specify the execution command, specify the execution command from the command information,
When the command information lacks information necessary to identify the execution command, an execution command identification that identifies the execution command using the command information and the information inferred by the lack information inference means. Means, a psychological distance sense inference means for inferring a degree of a psychological distance sense held by the user from the input sentence with respect to an object operated by the execution command, and the execution command identification means. Execution command, the matching degree inference means for inferring the degree to match the command intended by the input sentence, and when creating a response sentence corresponding to the execution command specified by the execution command identification means, At least one of the degree of the psychological distance sense inferred by the psychological distance inference inference means and the degree of coincidence inferred by the coincidence degree inference means is used. There is provided a command processing apparatus comprising: a response sentence creating unit that creates the response sentence by changing the expression of the response sentence; and a presentation unit that presents the response sentence created by the response sentence creation unit. (Third configuration).
The first configuration or the third configuration may further include history data of past command execution, and the shortage information inference unit may be configured to infer the shortage information using the command execution history data. (Fourth configuration). Further, the first configuration or the third configuration further includes a state obtaining unit that obtains a state of the target on which the execution command is executed at the time of obtaining the input sentence. The information obtained by the means may be configured to infer the insufficient information (fifth configuration). One of the first configuration or any of the third to fifth configurations has the most similar case sentence data corresponding to the command and the input sentence acquired by the input sentence acquisition means. A similar case sentence obtaining unit that obtains a case sentence from the case sentence data, wherein the coincidence inference obtaining unit uses the phrase of the input sentence and the degree of coincidence of the phrase of the case sentence to determine the degree of the match. (Sixth configuration). Further, the matching degree inference means of any one of the first or third to fifth configurations may be configured such that the inferred information inferring means infers missing information. Furthermore, the inference can be configured to infer the degree of coincidence using a correct probability (seventh configuration). Further, the matching degree inference means of any one of the first to third configurations or the third to fifth configurations includes: a case sentence data corresponding to the execution command; A similar case sentence obtaining means for obtaining, from the case sentence data, a case sentence most similar to the input sentence obtained in the step (a), and matching the phrase of the input sentence with the phrase of the case sentence obtained by the similar case sentence obtaining means Clause inference means for inferring the degree of match using the degree,
Probability coincidence inference means for inferring the degree of coincidence using the correct probability that the information inferred by the shortage information inference means,
A comparing unit that compares the degree of coincidence acquired by the phrase coincidence inferring unit with the degree of coincidence acquired by the probability coincidence inferring unit; and infers a large degree of coincidence by the comparing unit. It can be configured to be a value (eighth configuration). Further, from the first configuration or the third configuration to the eighth
In any one of the configurations up to the configuration of the above, the execution without presenting a response sentence to the user before the execution of the execution command, according to the degree of matching inferred by the matching degree inference means. It can be configured to execute a command (a ninth configuration). Further, the second configuration or the third configuration is
A psychological distance example sentence data composed using a psychological distance example sentence corresponding to a psychological distance sense held by the user with respect to the target operated by the execution command, and the input sentence acquiring means Further comprising: a psychological distance feeling case sentence acquiring unit that acquires the psychological distance feeling case sentence most similar to the acquired input sentence from the psychological distance feeling case sentence data; The psychological distance sense corresponding to the case sentence acquired by the psychological distance sense case sentence acquisition means may be configured as an inference value (first example).
0 configuration). Further, according to the present invention, in order to achieve the above object, a first step of obtaining an input sentence input in a natural language from a user, and command information from the input sentence obtained in the first step are obtained. A second step of acquiring, and, when the command information acquired in the second step lacks information necessary for specifying an execution command to be executed, infers the missing information. A third step to perform, and when the command information obtained in the second step includes information necessary to specify the execution command, the command is specified from the command information; If the information required to identify the execution command is insufficient, the execution command is identified using the command information and the information inferred in the third step. Step 4, a fifth step of inferring the degree to which the command specified in the fourth step matches the command intended by the input sentence, and an execution specified in the fourth step. A sixth step of creating a response sentence corresponding to the command by changing the expression of the response sentence according to the degree of matching inferred in the fifth step;
And a seventh step of presenting the response sentence created in the sixth step. Further, in order to achieve the above object, the present invention provides a first step of acquiring an input sentence input by a user in a natural language, and executing the first sentence obtained from the input sentence acquired in the first step. From the second step of acquiring command information used when specifying a command, and the input sentence acquired in the first step, the user holds the target operated by the execution command from the input sentence. A third step of inferring a degree of psychological distance feeling, and a step of creating a response sentence corresponding to the command specified by using the command information obtained in the second step, the third step A fourth step of changing the expression of the response sentence according to the degree of psychological distance inferred and a fifth step of presenting the response sentence created in the fourth step Provides a command processing method characterized in that it is composed of. Further, according to the present invention, in order to achieve the above object, an input sentence obtaining function for obtaining an input sentence input from a user in a natural language, and command information from the input sentence obtained by the input sentence obtaining function are provided. When the command information acquisition function to be acquired and the information required to specify the execution command to be executed in the command information acquired by the command information acquisition function are insufficient, the inferred information is inferred. Insufficient information inference function to be performed, and when the command information includes information necessary for specifying the execution command, the command is specified from the command information, and the execution command is specified in the command information. If the necessary information is insufficient, an execution command for specifying the execution command using the command information and the information inferred by the insufficient information inference function is used. A specifying function, a matching degree inference function for inferring a degree that the command specified by the execution command specifying function matches the command intended by the input sentence, and an execution command specified by the execution command specifying function When creating a response sentence corresponding to
A response sentence creation function for changing the expression of the response sentence according to the degree of matching inferred by the matching degree inference function, and a presentation function for presenting the response sentence created by the response sentence creation function And a computer-readable storage medium storing the command processing program for causing the computer to realize the above. Further, according to the present invention, in order to achieve the above object, an input sentence obtaining function for obtaining an input sentence input by a user in a natural language, and an execution from the input sentence obtained by the input sentence obtaining function, From the input information acquired by the command information acquisition function for acquiring the command information used when specifying the command and the input sentence acquisition function, the user holds the object operated by the execution command. When creating a response sentence corresponding to a command specified by using the command information acquired by the command information acquisition function and a psychological distance feeling inference function for inferring the degree of psychological distance feeling, the psychological distance A response sentence creation function for changing the expression of the response sentence according to the degree of psychological distance inferred by the feeling inference function, and a response sentence created by the response sentence creation function Command processing program for realizing a presentation function for presenting, to a computer a computer or storing the command processing program, provides a storage medium readable.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS In the present embodiment, a certainty factor indicating a degree of clarity of intention in a user's utterance and a sense of psychological distance held by the user with respect to the device are inferred from the content of the user's utterance. Thereby, the prior confirmation of the command execution is omitted, and the language used when interacting with the user is changed. Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to FIGS.

FIG. 1 is a diagram showing an example of the configuration of a command processing device 1 (sometimes called an agent) in the present embodiment. The microphone 9 is a part that acquires a natural language spoken by the user, such as “turn off the television” or “add a wavy line to this paragraph”.
In addition, the input device 10 is, for example, a keyboard, and the user can also input a natural language from the keyboard when performing a word processing operation or searching for help on a personal computer. Although the command processing device 1 includes both the microphone 9 and the other input devices 10, only one of them may be provided. The input unit 8 includes a bus line 3, a microphone 9, and other input devices 1.
This is a part that provides an interface with the “0”.

ROM (Read Only Memory)
y) 4, a basic program necessary for the operation of the command processing device 1 is recorded. RAM (Ran
A dom access memory (5) is a part that provides a working memory to the central processing unit 2 and loads various programs stored in the storage device 7 when executing them.

The storage device 7 is constituted by a hard disk or other non-volatile memory, etc., and includes a program unit 15 storing various programs and a database unit 16. The program unit 15 stores a command processing program for causing the command processing device 1 to perform command processing according to a predetermined algorithm, and other programs. The database 16 records a case dictionary used for analyzing a sentence input by a user, a state information database used for calculating a certainty factor described later, a command execution history, and the like.

The output section 11 is a section for providing an interface when the command processing apparatus 1 exchanges data with the outside. The output section 11 includes a device control section 12, a display 13, and a speaker 1
4 and so on. The device control unit 12 receives a command sent from the central processing unit 2 and turns off the television when the electronic device is a television, controls a floppy (registered trademark) disk drive when the electronic device is a word processor, and the like. , Where the command is executed. When the device is a word processor, the display 13 displays a message such as "make the background red," and provides a human interface for interacting with the user. The speaker 14 also issues a voice message such as “underline” to provide a human interface with the user.

The central processing unit 2 includes, for example, a CPU (Cen).
The processing unit is configured to control various arithmetic operations and each unit of the command processing device 1 according to a program stored in the ROM 4 or the program unit 15.

As described above, the command processing device 1
Is input from the input unit 8, the central processing unit 2 analyzes the input sentence according to the command processing program of the program unit 15 to obtain command information, and the command information is output from the output unit 11. . As described above, the command processing device 1 is configured to be compatible with a television and a word processor as an example, but is not limited thereto. For example, a keyboard is not required in the case of a television. The command processing program may be recorded in the ROM 4.

FIG. 2 is a diagram showing components constituting each function of the command processing apparatus 1.
Each component shown in FIG.
The command processing program stored in the RAM 5 is loaded into the RAM 5 and configured as software.
In this embodiment, each component is realized by software, but is not limited to this.
It may be configured by hardware.

The language analysis processing component 22 is a component that analyzes a natural language (command request utterance) obtained from the user and outputs sentence structure information. , How to form phrases and how those phrases are related. The confidence component 23 is a language analysis processing component 2
2, and outputs the certainty factor information and the execution command information to the response sentence creation component 25.

The certainty factor information is a numerical value from 0 to 1 indicating the degree to which the command intended by the user by the command request utterance matches the command specified by the command processing device 1 by supplementing the information lacking in the command request utterance. Things. Here, it is assumed that the degree of coincidence increases from 0 to 1. More specifically, the certainty factor is an index indicating the degree of certainty that the input sentence (command utterance) input by the user corresponds to the command information specified from the input sentence. is there. That is, when the command request utterance obtained from the user lacks information necessary for specifying the command, the command processing device 1 estimates the user's intention and specifies the command information for the time being. The less information that is missing, the higher the probability that the user's intention matches the command specified by the command processing device 1. The degree of certainty can be said to indicate the degree of information that is insufficient for executing the command. Thus, the certainty factor does not mean the user's confidence in executing the command.

The execution command information, when information required for command identification is missing in the user's command request utterance, supplements the information by guessing the missing information and representing the command identified from the complemented information. Information. Details of the certainty component 23 will be described later. The psychological distance component 24 infers the psychological distance that the user has in the device (hereinafter, the operated device) operated by the command processing device 1 from the ending expression of the user's command request utterance. Component. The psychological distance feeling component 24 sends the psychological distance feeling information obtained by quantifying the psychological distance feeling into a numerical value from −2 to +2 to the response sentence creating component 25. The psychological distance sense is a scale indicating whether or not the user's attitude toward the operated device is friendly. Details of the psychological distance component 24 will be described later.

The response sentence creating component 25 uses the certainty information acquired from the certainty component 23, the execution command information and the psychological distance information acquired from the psychological distance component 24 to respond to the user. A component that creates a sentence. The response sentence creation component 25 creates a response sentence 26 in which an utterance expression such as language is changed from the certainty factor information and the psychological distance information. Although not shown, the execution command information is shown in FIG.
Is sent to the output unit 11 and executed.

As the change of the utterance expression, for example, when changing the expression at the end of the sentence, as the degree of certainty increases from the smaller one, it is changed to “Water line”, “Let's make it a wavy line”, “Wave line”. ","Let's make it a wavy line, ""I'll make it a wavy line, "and" I make it a wavy line. " In addition, as the sense of psychological distance increases from the smaller one, the end of the sentence can be changed, such as "I like the wavy line,""Do you want to make it wavy?"

When a portion other than the end of a sentence is changed,
"Waves?", "Waves ..."
Expressions such as "Do you want to use wavy lines?", "What about wavy lines?", "What about wavy lines?", "Do you want to use wavy lines?" Can be changed.

Further, the method of confirming the command execution can be changed according to the certainty factor information. An example is shown below using a voice word processor in comparison with a conventional example.
The processing device indicates the command processing device 1. (Example of dialogue between conventional command processing device and user) User: draws a line. Processing unit: Execute "Dotted line". Is it OK? User: No, it's wavy. Processing unit: Executes "Make a wavy line". Is it OK? User: Yes. The operated device draws a wavy line after the response from the user.

(Example of dialogue between the command processing device of the present embodiment and the user) User: draw a line → small confidence processing device: why not use a "dotted line"? User: No, it's wavy. → Confidence: Large The operated device draws a wavy line after the user's response. Processing device: "wavy line".

Since the certainty component 23 has determined that the certainty of the user's utterance of "drawing a line" is small, the processing apparatus asks the user to confirm the command before executing the command, "Let's try a dotted line?" Do. Further, since the certainty component 23 has determined that the certainty factor of the user's utterance of “No, to make a wavy line” is large, the processing device executes “to make a wavy line.” By omitting the confirmation before executing the command, the operated device immediately draws a wavy line. Then, the command processing device 1 subsequently reports to the user after the command execution.

As described above, by changing the method of confirming command execution according to the degree of certainty, if the user's intention is sufficiently understood, it is possible to perform post-report utterance without performing pre-command execution confirmation each time. The load on the user is reduced. As described above, the response sentence creation component 25 changes the expression method for expressing the command information obtained as the execution command information to the user based on the certainty factor information and the psychological distance information.

FIG. 3 is a diagram showing the structure of the certainty component 23. The outline of this component is as follows. The certainty component 23 calculates a first certainty from a comparison between the input sentence and the case sentence in the case dictionary 33, obtains a second certainty from a comparison between the input sentence and the state of the operated device, and Third from statement and past command execution history
Is calculated, and the highest certainty factor among these certainty factors is output as certainty information 40. If there is missing information in the input sentence, the execution command information 41 is output by guessing and complementing the missing information from the past command execution history.

The details of the certainty component 23 will be described below. The sentence structure information 31 sent from the language analysis processing component 22 in FIG. 2 is input to the case matching processing unit 32. The case collation processing unit 32 stores the case dictionary 3
3 is compared with the user's command request utterance, and execution command information associated with the case sentence that is considered to be most similar is acquired.

In the present embodiment, the first certainty factor is defined by the following equation (1) as an example. (Confidence) = (number of clauses matching case sentence) / (number of clauses of case sentence) (1) The case matching processing unit 32 sends the first certainty to the certainty information determination unit 39. At the same time, the execution command information is output to the missing information inference unit 34. This execution command information may be partially missing. Case matching processing unit 3
A specific example of the process performed by No. 2 will be described later.

The missing information inference unit 34 is provided with the state information inference unit 3
7 and an execution history inference unit 38. The state information inference unit 37 acquires the state information of the operated device (for details, for example, information that the receivable channels are 1, 3, and 12 in the case of a television, which will be described later), and performs case comparison based on this information. The missing part is estimated from the execution command information sent from the unit 32, and a second certainty factor is calculated based on the estimation. The missing information inference unit 34 creates execution command information by complementing the missing information. The state information inference unit 37 sends the second certainty factor to the certainty factor information determination unit 39.

In this embodiment, the state information inference unit 37
Calculates the second certainty factor under the following conditions. When it is not necessary to infer missing information using state information, the second certainty factor is set to 1.0. When inference of missing information is performed using state information, when only one answer is obtained, the second missing information is set to 0.9. When inferring missing information using state information, and when a plurality of answer candidates are obtained, it is assumed that 0.9 is divided by the number of candidates.
Note that “when it is not necessary to infer missing information using state information” means that the case matching processing unit 32 uses the sentence structure information 31
Means that there is no missing information in the sentence structure information 31. Specific examples of the processing performed by the missing information inference unit 34, including the execution history inference unit 38 described below, will be described later.

The execution history inference unit 38 receives, from the execution history database 36, a history related to an input given to the operated device by the user in the past and a history related to a command executed by the operated device (the two histories are combined). Based on these histories, information missing from the execution command information sent from the case matching processing unit 32 is inferred. Then, the execution command is complemented and output based on the inferred missing information. Further, a third certainty factor is calculated at the time of inference, and the certainty factor information determining unit 3
9

In the present embodiment, as an example, the third certainty factor is calculated by the execution history inference unit 38 under the following conditions. If inference from the execution history is not required, the third certainty factor is set to 1.0. When inference is performed from the execution history, the third certainty factor is set to a value obtained by multiplying the probability that the inference result is correct by 0.6. Here, the probability that the inference result is correct is obtained from the execution history. A specific example of the processing performed by the execution history inference unit 38 will be described later. Note that “when there is no need to infer missing information using state information” means that the sentence structure information 31 and the execution history are compared by the execution history inference unit 38, and that there is no missing information in the sentence structure information 31. Means that

The certainty determining unit 39 determines the first certainty, the second certainty,
, And a third certainty, and compare their magnitudes. Then, the one with the highest certainty is sent to the response sentence creating component 25 in FIG. 2 as certainty information. If the second credibility is greater than the third conviction as a result of the comparison in the credibility determination section 39, the missing information inference section 34 adds the execution command information complemented by the state information inference section 37 to the response statement. When the third certainty is greater than the second certainty, the execution history inference unit 38 sends the completed execution command information to the response sentence creation component 25.

Hereinafter, a case where the certainty component 23 obtains certainty information and execution command information will be described using a specific example. (Specific Example 1) A television is assumed as the operated device. In this case, it is assumed that the user gives an input of “I want to watch TV” to the command processing device 1. This input is analyzed by the language analysis processing component 22 of FIG. 2, and the analysis result is input to the certainty factor component 23 as sentence structure information.

The case collating unit 32 operates as follows. The case matching processing unit 32 searches the case dictionary 33 for a case considered to be most similar. As a result, [[number] /
It is assumed that the case of "I want to watch / TV on the channel" is analyzed to be the most similar. However, phrases are separated by / symbols. Since the number of clauses that match the case sentence (four clauses) among the clauses of the sentence input as the sentence structure information is 2 (“I want to see” and “TV”), the confidence is 2/4 = 0.
It becomes 5.

As a result of the case collation processing in the case collation processing unit 32, the channel information is missing, so that the state information inference unit 37 performs the state information inference processing. It is assumed that three channels 1, 3, and 12 can be viewed on the television. This information is recorded in the state information database 35. At this time, the answer candidates obtained by the state information inference unit 37 and their certainty factors are as shown in FIG. That is, there are three types of answer candidates: “watch 1-channel TV”, “watch 3 channel TV”, and “watch 12 channel TV”, each having the same certainty and the certainty from 0. Since it takes a value up to 1, the certainty factor of each answer candidate is 1/3 = 0.3.
However, figures after the second decimal place have been rounded down. in this way,
If the certainty of each answer candidate is equal, the state information inference unit 37
For example, when there are three candidates, the candidates are ranked, and the missing information is complemented by adopting a candidate having a higher rank, or by suggesting three candidates to the user.

The execution history inference unit 38 uses the past execution history of the execution history database 36 based on the day of the week or the time period when the user made a command request, and displays the channel of the channel that the user was watching during the day of the week and the time period. Analyze trends. For example, it is assumed that the command request is made on Monday night at 9:00. As a result of examining the past execution history of the execution history database 36 by the execution history inference unit 38, as shown in FIG. 5% for 3 channels, 12
Assume that the channel is 5%. At this time, as shown in FIG. 4C, the certainty factor of each answer candidate is 0.54 for “watch 1-channel TV”, 0.03 for “watch 3-channel TV”, and “12 channel”. Watch TV "is 0.03. The supplemented execution command information is “Watch one-channel television”.

Before explaining the second embodiment, the third confidence level is adjusted to define the reliability level for improving the accuracy. The reliability indicates the degree to which the user trusts the command processing device 1-
A value from 2 to +2, which is the third
In determining the value of the reliability, the final calculated value is adjusted up or down. If the user determines that the analysis result of the command processing device 1 is trusted, the third confidence factor is corrected to a larger value by the correction based on the reliability, and the user checks the analysis result of the command processing device 1. If it is determined that the user does not trust much, the third certainty factor is corrected to a smaller value by the correction based on the reliability.

The value of the reliability is obtained by the following factors. Expressions of reprimand or abuse such as "What are you doing?", "Is different", or "Isn't it?" When the operated device executes a command that is different from the user's expectation Is reduced (for example, the reliability is set to −1), the command processing device 1 decreases the reliability. On the other hand, from the user, "great", "do it", "can do",
"Yes" (for example, the reliability is set to +1)
When the command processing device 1 acquires an affirmative / praise expression for the command execution result, the reliability increases.

Further, the following words can be considered as words that the user can trust and determine the command processing device 1. When both the speaker and the listener are familiar with the instruction target, the instruction pronouns “Are” and “Ano” are used without using a specific instruction name. Also, the assumption that "you always know what to do because you use it" is sometimes called "usual". Therefore, the user uses these demonstrative pronouns without using a specific name, for example,
When a command request such as “that file” or “want to see that file” is made to the command processing device 1 (for example, the reliability is set to +2), the reliability is increased.

The third certainty factor can be defined as in the following equation (2) using reliability as an example. (Third reliability) = (probability that the obtained result is correct) × (0.6 + 0.15 × (reliability)) (2)

(Specific Example 2) In specific example 2, as shown in the equation (2), the third certainty factor is obtained in consideration of the reliability factor. It is assumed that the user gives an input of “I want to see the usual” to the command processing device 1 installed in the operated device (the television in this specific example). Since the phrase “usual” that can determine that the user trusts the command processing device 1 is included, the reliability is +2. Based on the day of the week and the time zone when the user made the command request, the execution history inference unit 38 uses the past execution history recorded in the execution history database 36 to determine the tendency of the channel that the user saw on the day and the time zone. analyse.

As a result, FIG. 5 (a) shows the ratio of the tendency of the channel that the user has watched in the corresponding time zone of the day of the week, obtained as a result of the analysis by the execution history inference unit 38. In the past, the percentage of channels that the user had watched during this time of the day on this day was 70% for 1 channel and 25% for 3 channels.
%, 12 channels were 5%. The case where the third certainty is not corrected using the reliability (the calculation method in the specific example 1), and the ratio of the reliability (+2) and the ratio (the probability that the user intentionally requests the command for the channel and the estimation) FIG. 5 (b) shows a value obtained when the third certainty factor is corrected by substituting the above-mentioned value into Expression (2).

As a result, the third certainty factor for the answer candidate “I want to watch one-channel television” is 0.42 when not corrected by the reliability, and 0.63 when corrected by the reliability. And the degree of certainty increases. Conversely, when the reliability becomes a negative value, for example, -1, the third certainty corrected by the reliability becomes smaller than the third certainty without correction. in this way,
The third certainty factor, that is, the user's certainty factor inferred from the past command execution history,
The accuracy of the inference can be increased by modifying the reliability using the reliability held for.

FIG. 6 is a flowchart showing a processing procedure performed by the certainty component 23. First, sentence structure information is acquired from the language analysis processing component 22 of FIG. 2 (step 10). Next, in the case matching processing unit 32,
The input sentence obtained from the sentence structure information is compared with the most similar sentence from the case sentences in the case dictionary 33 to determine the presence or absence of missing information, and then the execution command information is sent to the state information inference unit 37 (step 20). Further, the case matching processing unit 32 obtains the first certainty factor by comparing the input sentence and the case sentence, and sends the first certainty factor to the certainty factor information determining unit 39 (step 3).
0). Next, the state information inference unit 37 acquires the state information of the operated device from the state information database 35, estimates and complements missing information for the execution command information that has been case-matched (step 40), and completes the completion. The execution command information 1 is created (step 50), and the second certainty is sent to the certainty information determining unit 39 (step; 60).

Next, the execution history inference unit 38 uses the execution history of the execution history database 36 to execute the case matching processing unit 32.
An execution history inference process is performed on the execution command information transmitted from the server (step 70). Execution history inference unit 38
Creates the completed execution command information 2 by guessing and complementing the information missing in the execution command information (step 8)
0). In addition, a third certainty factor is calculated and sent to the certainty factor information determining unit 39 (step 90).

Next, the certainty information determiner 39 compares the respective certainties with each other. If the first certainty is the largest (step 100; Y), the first certainty is returned as the certainty information. (Step 16)
0). The certainty factor determination unit 39 further compares the magnitudes of the second and third certainty factors (step 170). If the second certainty factor is larger than the third certainty factor (step 170; Y),
The missing information inference unit 34 sends the execution command information 1 to the response sentence creation component 25 (step 150). When the third certainty is larger than the second certainty (Step 170; N), the missing information inference unit 34 sends the execution command information 2 to the response sentence creation component 25 (Step 130).

On the other hand, as a result of the comparison by the certainty factor information decision unit 39, the second certainty factor is the highest (step 110;
Y), the certainty factor information determination unit 39 sends the second certainty factor to the response sentence creation component 25 (step 140),
The missing information inference unit 34 sends the execution command information 1 to the response sentence creation component 25 (Step 150).

If the third certainty is the largest as a result of the comparison by the certainty information determining unit 39 (step 110; N),
The certainty information determination unit 39 sends the third certainty to the response sentence creation component 25 (step 120), and the missing information inference unit 34 sends the execution command information 2 to the response sentence creation component 25 (step 130). ).

Next, the function of the psychological distance component 24 shown in FIG. 2 will be described. The psychological distance component 24 uses the end of a verb included in the user's utterance (a sentence that the user has uttered to the command processing device 1), and the like, and the psychological sense of the user with respect to the command processing device 1. Quantitative sense of distance. In the present embodiment, the standard psychological distance sense is represented by 0, the distant psychological distance sense is represented by -2, and the intimate psychological distance sense is represented by +2.
The sense of psychological distance was divided into five levels by integers from to +2.

Then, for example, information that the psychological distance of the user to the command processing apparatus 1 is smaller in the informal expression of “to do it” than in the polite request expression of “can you do it?” Numerical values are stored in a table. This table is stored in the database unit 16 of FIG. Then, when the user requests a command, the corresponding expression is searched to quantify the psychological distance sensed by the user.

FIG. 7 shows an example of a table in which the feeling of psychological distance held by the user with respect to the command processing apparatus 1 is quantified, paying attention to the ending expression of the verb in the request expression. When the end expression of the user's command request utterance is “can you do it?”, The numerical value of the psychological distance held by the user in the command processing device 1 is the largest, and this is the psychological distance. Feeling component 2
Numerical conversion with 4 results in a minimum of -2. Similarly, when the end expression is "Please do" or "Would you like to do", the numerically sensed psychological distance is -1.

Similarly, the ending expression is “to want” or “to
When "do", the value obtained by digitizing the psychological distance is 0. The numerical value of the psychological distance feeling increases as the tail expression becomes “to do” and “to do”, and becomes +1 and +2, respectively.

Hereinafter, as an example, a specific processing procedure of the psychological distance component 24 will be described using the command processing device 1 mounted on a television. If the user's command request utterance is “Can you turn off the television?”, The psychological distance component 24 searches the table shown in FIG. Is determined to match the end expression. Next, the psychological distance component 24 obtains a value -1 obtained by digitizing the psychological distance feeling corresponding to the end expression. Then, this numerical value is output to the response sentence creation component 25. Similarly, when the user's command request utterance is “turn off TV”, the verb end expression matches “to do” in the table of FIG. 7, and the value obtained by numerically expressing the sense of psychological distance is It becomes +1.

The psychological distance sensed by the user with respect to the command processing device 1 can be determined by the words other than the end expression of the verb, if the words can be used to determine the psychological distance sensed by the user. Can be quantified. For example, before the user's request expression,
・ ”,“ I'm sorry ... ”, etc., and the numerical value is set to −1, for example, because the psychological sense of distance is sparse.

When the nickname given to the command processing device 1 by the user is "Rabbit" and the user calls on the command processing device 1 "Mr. Rabbit", the value of the psychological distance is set to +1. , "Rabbit-chan" and "Rabbit-kun" are +2, and the sense of psychological distance can also be quantified by the words of the user's call to the command processing device 1.

Next, the response sentence creation component 2 shown in FIG.
5 describes a method of creating the response sentence 26 using the certainty factor information, the execution command information, and the psychological distance feeling information. The utterance content is determined by the execution command information, and the method of expressing the utterance content is determined from the certainty factor information and the psychological distance information. More specifically, the utterance content is determined by searching the utterance content corresponding to the execution command information, similarly to the command processing device of the conventional example. The method of expression is determined based on the certainty factor information and the psychological distance information based on the obtained utterance content, whether to perform the pre-command execution confirmation or only the post-report, and to determine the utterance content. Determine changes in endings and phrases.

A specific example in which the response sentence creation component 25 determines the utterance content and the expression method will be described. In order to determine an expression method from the certainty factor information and the psychological distance feeling information, for example, a table as shown in FIG.
Is stored in the database unit 16. Here, the four levels of confidence shown in the column of confidence are “low”, “medium”, “high”,
“Highest” was classified as follows as an example. Maximum ... (confidence) = 1.0 High ... 0.7≤ (confidence) <1.0 Medium ... 0.4≤ (confidence) <0.7 Low ... (confidence) Degree) <0.4

As can be seen from the table shown in FIG. 8, in this specific example, the response sentence creation component 25 includes a value representing the psychological distance sense obtained from the psychological distance sense information and a confidence value obtained from the certainty factor information. The response sentence 26 is created from a combination of values representing degrees. For example, if the confidence is medium and the psychological distance is 0, the end of the response sentence 26 is “Do you want to do?”. If the confidence is “low” and the psychological distance is +2, The end of the sentence 26 is "Try it out." Similarly for other combinations of certainty and psychological distance, the tail expression of the response sentence is changed according to the table of FIG.

The response sentence 26 is used as a confirmation utterance before the command execution for the user or a post-report utterance after the command execution. If the confidence is low, medium,
The response sentence 26 of “high” is used as a pre-command execution confirmation utterance, and the response sentence 26 of which the confidence is “highest” is used as a post-report utterance.

According to the table shown in FIG. 8, the response sentence creation component 25 not only creates the response sentence 26 at the same level as the psychological distance sense acquired from the user utterance, but also creates the response sentence 26 acquired from the user utterance. The response sentence 26 can be created at a level different from the sense of distance. Answer sentence 2 at the same level as the psychological distance sense obtained from the user utterance
6 is as follows. When the user issues a command request to "turn off the television", the utterance content is "turn off the television", the certainty factor is "highest", and the psychological distance is -1. The response sentence 26 corresponds to FIG.
According to the table, the ex-post report utterance "Turn off the TV" is made.

On the other hand, when the response sentence 26 is created at a level different from the psychological distance sense acquired from the user's utterance, for example, the following can be performed. When the user requests the TV to “turn off”, the utterance content is “turn off”, the confidence is “low”, the psychological distance is +2, but the psychological distance corresponds to −1. Make a humble expression such as "How about turning off the TV?"

Next, four examples of interaction between the command processing device 1 configured as described above and the user will be described together with an example of interaction with a conventional command processing device, taking a case where the command processing device 1 is implemented in a voice word processor as an example. In addition, “→” in the notation indicates a certainty factor, a sense of psychological distance, or the like acquired by the command processing device 1 when the command request is made. First, as a usage example of the certainty factor, a dialogue example when the user's certainty factor is high will be described. (Example of dialogue between conventional command processing device and user) User: I want to draw a wavy line. → Confidence: Highest Processing unit: Execute "Draw a wavy line". Is it OK? User: Yes. The operated device (voice word processor) draws a wavy line after the response from the user. (Example of Interaction between User and Command Processing Device of Present Embodiment) User: I want to draw a wavy line. → Confidence: Highest The operated device (voice word processor) draws a wavy line. Processing equipment: The wavy lines are drawn. In this example of the dialogue, since the certainty degree of the utterance content of the user is the highest, the command processing device 1 does not perform the pre-command execution confirmation, but does the post-report.

Next, as an example of the use of the certainty factor, an example of the dialogue when the certainty factor is low will be described. (Example of Dialogue between Conventional Command Processing Device and User) User: Want to stand out. → Low confidence: Processing unit: Which command should be executed? Thereafter, the operated device (voice word processor) displays a dialog of command candidates. (In the case where the command processing apparatus of the present embodiment and the user show a candidate with the highest certainty factor in a dialogue example) User: Want to stand out. → Confidence: Low Processing unit: Why don't you enlarge the characters? (In the case of a dialogue example between the command processing device and the user according to the present embodiment, some candidates having a high degree of certainty are exemplarily shown.) User: Want to stand out. → Confidence: Low Processing unit: Try to enlarge the text, add color, or draw a line.

Next, as an example of using the sense of psychological distance, an example of dialogue when the sense of psychological distance is small will be described. (Example of dialogue between the conventional command processing device and the user) User: Rabbit. If it's a wave line. → Psychological sense of distance: +2 Processing unit: Executes “draw a wavy line”. Is it OK? User: Yeah, good. Thereafter, the operated device (voice word processor) draws a wavy line.
(Example of dialogue between the command processing device of this embodiment and the user) User: Rabbit. Even if it's a wavy line. → Psychological sense of distance: +2 Processing unit: Can I draw a wavy line? User: Yeah. OK. Thereafter, the operated device (voice word processor) draws a wavy line.

Next, an example of the dialogue in the case where the confidence increases due to the reliability obtained from the user's command request utterance will be described. It is assumed that the user normally draws a wavy line as a means for making the document stand out. (Example of interaction between user and command processing device of the present embodiment) User: I want to make this stand out. → Confidence: Low Processing unit: Why don't you enlarge the characters? →
Confidence: Low User: Do it the usual way. → Reliability: +2
(This increases confidence.) Processor: Then, draw a wavy line. → Confidence: High User: Yeah. Please. Thereafter, the operated device (voice word processor) draws a wavy line.

In the above example of a voice word processor, the sense of psychological distance was determined from the content of the user's command request utterance. However, the present invention is not limited to this. When selected, the command processing device 1
Can be configured so that the user can select such that the user always responds with a sense of psychological distance of “−2”.
This can be applied, for example, to an online shopping server that interacts with customers who require polite language. Hereinafter, an example in which the user interacts with the command processing device 1 mounted on the online shopping server will be described. User: I want to see a sweater. Processing equipment: What kind of sweater do you have? → Psychological distance: -2 User: Men's, wool-made material is good. Processing equipment: How about this one? → Psychological distance: -2 After this, a picture of the sweater is displayed on the user's terminal device by the net shopping server. User: Hmm. I like blue instead of red. Processing equipment: Blue. Then how about this one? → Psychological distance: -2 After this, another sweater picture is displayed on the user's terminal device by the net shopping server.

FIG. 9 is a flowchart showing the operation procedure of the command processing device 1. First, the command processing device 1 acquires a user's command request utterance (input sentence) via an external input device such as a microphone 9 or a keyboard (FIG. 1) connected to the command processing device 1 (step 2).
10). Next, the language analysis processing component 22 shown in FIG. 2 analyzes the input sentence and acquires sentence structure information. Next, the sentence structure information is input to the certainty component 23 and the psychological distance component 24.

The certainty component 23 acquires execution command information from the sentence structure information (step 2).
30), obtain certainty factor information (step 240). On the other hand, the psychological distance component 24 acquires psychological distance information from the sentence structure information (step 250). Next, the response sentence creation component 25 acquires the certainty factor information and the execution command information from the certainty factor component 23, and acquires the psychological distance sense information from the psychological sense of distance component 24, and based on these information. Response 26
Is created (260).

In the present embodiment described above, the following effects can be obtained. First, by using the certainty factor, it is possible to determine whether or not it is necessary to perform a pre-command execution check, and it is possible to reduce stress applied to the user. Further, even when the degree of certainty obtained from the user's command request utterance is low, it is possible to propose a command by dialogue, for example, "Would you like to change the background to blue?" In addition, "Do as usual"
For example, even if the user does not specifically convey the target command, it is possible to respond by inferring the user's intention from the past command execution history.

In addition, by digitizing the feeling of psychological distance from the user's command request utterance, the command processing device 1 can return a response sentence having various sentence end expressions and nuances accordingly. This eliminates the need for monotonous and mechanical dialogue. For example, a user who desires a dialogue with a large numerical value representing a sense of psychological distance, that is, a user who has a small sense of psychological distance with the operated device, can perform a more friendly conversation. Further, even when information for command execution is insufficient in the command request utterance from the user, it is possible to respond by estimating the command intended by the user from the past command execution history or the like.

In the present embodiment, the command processing device 1 is implemented as an example in a television and an audio word processor. However, the present invention is not limited to this. For example, other home appliances such as a microwave oven, a rice cooker, and a washing machine may be used. And CAD (Computing
The present invention can also be applied to other computer software such as er Aided Design. Furthermore,
The command processing device 1 can be mounted on a server and used for interaction between the server and a user via a network such as the Internet. It can also be implemented in a digital television to provide a human interface for performing bidirectional communication between the digital television and a carrier (such as a broadcasting station).

The command processing device 1 may be, for example, a bed, a robot, a car, or the like. For example, a bed equipped with the command processing device 1 can change the angle of the surface of the bed in response to a user's command request utterance, and is effective for a nursing bed or the like. Further, the robot equipped with the command processing device 1 can be operated by a user's command request utterance. Also, in a vehicle equipped with the command processing device 1, a user conventionally manually performs, for example, turning on / off a headlight, operating a wiper, starting and stopping an engine, operating an air conditioner, and the like, according to a user's command request utterance. Operations can be performed.

[0079]

According to the present invention, a command processing device capable of accurately grasping a user's request, reducing a user's input burden, and enabling a dialog taking into account a psychological distance held by the user with respect to the device. , A command processing method, and a command processing program.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of a configuration of a command processing device according to the present embodiment.

FIG. 2 is a diagram showing components constituting each function of the command processing device.

FIG. 3 is a diagram showing a configuration of a certainty factor component.

FIG. 4 is a diagram showing answer candidates obtained by a state information inference unit and their certainty factors;

FIG. 5 is a diagram showing, for example, a ratio of a tendency of a channel watched by the user during the time period on the day of the week, obtained by the execution history inference unit as a result of the analysis.

FIG. 6 is a flowchart illustrating a processing procedure performed by a certainty component.

FIG. 7 shows an example of a table in which a sense of psychological distance is quantified.

FIG. 8 is a diagram showing an example of a table for determining an utterance content and an expression method.

FIG. 9 is a flowchart showing a procedure of an operation of the command processing device.

FIG. 10 is a diagram showing an example of a configuration of a conventional IIF analysis engine.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 Command processing device 2 Central processing unit 3 Bus line 4 RAM 5 ROM 7 Storage device 8 Input unit 9 Microphone 10 Other input device 11 Output unit 12 Device control unit 13 Display 14 Speaker 15 Program unit 16 Database unit 21 Input sentence 22 Language analysis Processing component 23 Confidence component 24 Psychological sense of distance component 25 Response sentence creation component 25 26 Response sentence 31 Sentence structure information 32 Case matching processing unit 33 Case dictionary 34 Missing information inference unit 35 State information database 36 Execution history database 37 State information inference Unit 38 execution history inference unit 39 certainty information determining unit 40 certainty information 41 execution command information

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 15/28 G10L 3/00 571Z

Claims (14)

[Claims] An input sentence obtaining unit that obtains an input sentence input in a natural language from a user; a command information obtaining unit that obtains command information from the input sentence obtained by the input sentence obtaining unit; When information necessary for specifying an execution command to be executed in the command information acquired by the command information acquisition means is insufficient, insufficient information inference means for inferring the insufficient information; and the command When the information includes information necessary for specifying the execution command, the command is specified from the command information, and the command information lacks information necessary for specifying the execution command. If
An execution command identification unit that identifies the execution command using the command information and the information inferred by the shortage information inference unit; and a command identified by the execution command identification unit,
A matching degree inferring means for inferring a degree matching the command intended by the input sentence, and when creating a response sentence corresponding to the execution command specified by the execution command specifying means, the matching degree inferring means Response sentence creating means for creating the response sentence by changing the expression of the response sentence according to the degree of coincidence inferred, and presentation means for presenting the response sentence created by the response sentence creation means. A command processing device characterized by the above-mentioned. 2. An input sentence obtaining means for obtaining an input sentence input by a user in a natural language, and command information used for specifying an execution command from the input sentence obtained by the input sentence obtaining means. From the input sentence acquired by the input sentence acquiring means, and infers the degree of psychological distance the user has with respect to the target operated by the execution command. A psychological distance feeling inference means, and a psychological inference by the psychological distance feeling inference means when creating a response sentence corresponding to the command specified by using the command information obtained by the command information obtaining means. Response sentence creating means for creating the response sentence by changing the expression of the response sentence according to the degree of a sense of distance, and presentation means for presenting the response sentence created by the response sentence creating means. Command processing apparatus characterized by the. 3. An input sentence obtaining means for obtaining an input sentence input in a natural language from a user; a command information obtaining means for obtaining command information from the input sentence obtained by the input sentence obtaining means; When information necessary for specifying an execution command to be executed in the command information acquired by the command information acquisition means is insufficient, insufficient information inference means for inferring the insufficient information; and the command When the information includes information necessary to specify the execution command, the execution command is specified from the command information, and the command information lacks the information necessary to specify the execution command. If so, execution command specifying means for specifying the execution command using the command information and the information inferred by the shortage information inference means; A psychological distance feeling inference means for inferring a degree of a psychological distance feeling held by the user with respect to an object operated by the execution command, and the execution command specified by the execution command specifying means, A matching degree inferring means for inferring a degree of matching with a command intended by an input sentence; and, when creating a response sentence corresponding to the execution command specified by the execution command specifying means, Response sentence creating means for creating by changing the expression of the response sentence by using at least one of the degree of psychological distance sense inferred at and the degree of coincidence inferred by the coincidence inference means, A command processing device, comprising: a presentation unit that presents a response sentence created by a response sentence creation unit. 4. The apparatus according to claim 1, further comprising history data of past command execution, wherein said shortage information inference means infers the shortage information using the history data of command execution. 3. The command processing device according to 3. 5. An execution target of the execution command,
The information processing apparatus according to claim 1, further comprising a state acquisition unit that acquires a state at the time of acquiring the input sentence, wherein the shortage information inference unit infers the insufficient information using the state acquired by the state acquisition unit. The command processing device according to claim 1 or 3. 6. A case sentence data corresponding to a command, and a similar case sentence obtaining unit for obtaining a case sentence most similar to the input sentence obtained by the input sentence obtaining unit from the case sentence data, 6. The method according to claim 1, wherein the matching degree obtaining unit infers the degree of matching using a matching degree between a phrase of the input sentence and a phrase of the case sentence. 7.
The command processing device according to claim 1, wherein 7. The coincidence inference means inferring the degree of coincidence using a correct probability when the inference information is inferred by the insufficiency information inference means. The command processing device according to claim 1, wherein the command processing device is any one of claims 1 to 3. 8. The similarity inference means, wherein: a case sentence data corresponding to the execution command; and a similar case acquiring from the case sentence data a case sentence most similar to the input sentence acquired by the input sentence acquiring means. Sentence acquisition means, clause match degree inference means for inferring the degree of match when using the degree of match between the phrase of the input sentence and the phrase of the case sentence acquired by the similar case sentence acquisition means, and the lack information Probability coincidence inference means for inferring the degree of coincidence using the correct probability of the information inferred by the inference means; the coincidence acquired by the phrase coincidence inference means; And comparing means for comparing the degree of coincidence, wherein a value of the degree of coincidence in the comparing means is set as an inferred value. 1 request for any of The command processing device according to the item. 9. The execution command is executed without presenting a response sentence to the user before the execution of the execution command, according to the degree of matching inferred by the matching degree inference means. Claim 1 and Claims 3 to 8
A command processing device according to any one of the preceding claims. 10. A sentence of psychological distance example sentence data configured using a psychological distance feeling example sentence corresponding to a psychological distance feeling held by a user with respect to an object operated by the execution command; Further comprising: a psychological distance sense case sentence obtaining unit that obtains the psychological distance sense case sentence most similar to the input sentence obtained by the sentence obtaining unit from the psychological distance sense case sentence data; 4. The feeling inference means, wherein the psychological distance feeling corresponding to the case sentence acquired by the psychological distance feeling example sentence acquisition means is used as an inference value.
2. A command processing device according to claim 1. 11. A first step of obtaining an input sentence input by a user in a natural language, a second step of obtaining command information from the input sentence obtained in the first step, A third step of inferring the missing information when the information necessary for specifying the execution command to be executed is insufficient in the command information acquired in the second step; If the command information obtained in step 2 includes information necessary for specifying the execution command, the command information is specified from the command information, and the command information is specified in the command information. A fourth step of identifying the execution command using the command information and the information inferred in the third step when the necessary information is insufficient; A fifth step of inferring the degree to which the command specified in the step corresponds to the command intended by the input sentence, and a response sentence corresponding to the execution command specified in the fourth step In doing so, a sixth step of changing the expression of the response sentence according to the degree of matching inferred in the fifth step, and creating the response sentence created in the sixth step 7. A command processing method, comprising: presenting a seventh step; 12. A first step of obtaining an input sentence input in a natural language from a user, and command information used to specify an execution command from the input sentence obtained in the first step From the input sentence obtained in the first step, and infers a degree of psychological distance sensed by the user with respect to an object operated by the execution command. A third step, and when creating a response sentence corresponding to the command specified by using the command information acquired in the second step, the sense of psychological distance inferred in the third step is A fourth step of changing the expression of the response sentence according to the degree, and a fifth step of presenting the response sentence created in the fourth step. Command processing method and butterflies. 13. An input sentence obtaining function for obtaining an input sentence input in a natural language from a user; a command information obtaining function for obtaining command information from the input sentence obtained by the input sentence obtaining function; When information necessary to specify an execution command to be executed in the command information acquired by the command information acquisition function is insufficient, an insufficient information inference function for inferring the insufficient information, and the command When the information includes information necessary for specifying the execution command, the command is specified from the command information, and the command information lacks information necessary for specifying the execution command. If
An execution command identification function that identifies the execution command using the command information and the information inferred by the shortage information inference function, and a command identified by the execution command identification function,
A matching degree inference function for inferring the degree of matching with the command intended by the input sentence, and when creating a response sentence corresponding to the execution command specified by the execution command specifying function, the matching degree inference function A response sentence creation function for creating the response sentence by changing the expression of the response sentence according to the degree of the match inferred by the computer; Command processing program for causing 14. An input sentence obtaining function for obtaining an input sentence input in a natural language from a user, and command information used for specifying an execution command from the input sentence obtained by the input sentence obtaining function. From the input sentence acquired by the command sentence acquisition function obtained by the input sentence acquisition function, and infers the degree of the psychological distance sensed by the user with respect to the target operated by the execution command. When creating a response sentence corresponding to a command specified by using the command information acquired by the command information acquisition function, the psychological inference inferred by the psychological distance inference function. A response sentence creation function that creates the response sentence by changing the expression of the response sentence according to the degree of the sense of distance, and a presentation function that presents the response sentence created by the response sentence creation function. Command processing program for realizing a computer.