JP2003330920A - Data accumulation type conversation device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To allow a user to enjoy a natural conversation while recalling a past story. <P>SOLUTION: Conversation data are memorized by assigning a rank according to the level of importance and frequency of words. The conversation data having a low rank are extinguished, but conversation data having a rank higher than a certain level remain. Past conversation data are inserted between conversations. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot for conversation. In particular, it relates to robots that have a natural dialogue by making use of what they have talked about in the past.

[0002]

2. Description of the Related Art Conventionally, there have been robots that have a conversation based on built-in data. Even if the content to be spoken is human, the pattern of knowledge and the story is a combination of few elements, and the structure of the story is simple, and the mechanical impression is inevitable.

[0003]

DISCLOSURE OF THE INVENTION In the present invention, the above-mentioned drawbacks are solved, the contents spoken by a human being are stored, past stories are reproduced according to the importance of the topic, and the present conversation is reproduced. By interposing, I realized a conversation as if I was talking to a human being very naturally.

Furthermore, the contents of the story are erased in the order of importance. In addition, I tried not to delete the most important contents of the story.

[0005]

In order to achieve the above object, a data storage type conversation device of the present invention comprises a conversation sentence input means for inputting a conversation sentence, a conversation sentence storing means for storing the conversation sentence,
Conversational sentence output means for outputting the conversational sentence, conversational sentence ranking means for ranking the conversational sentence, and conversational sentence erasing determination for determining whether or not to erase the conversational sentence ranked by the conversational sentence ranking means. Means for erasing the conversational sentence by the conversational sentence erasure judging means.

Furthermore, as a basis of discrimination by the conversation sentence erasure discrimination means, the importance or frequency of use of words used in the conversation sentence is used.

Furthermore, the output timing of the conversational sentence output means depends on the type of the input word by the conversational sentence input means.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings.

(Outline) First, an outline of the operation of this apparatus will be described. FIG. 1 shows a conceptual diagram of data and the flow of data, which will be described with reference to this.

First, it is assumed that a conversation (101) between the apparatus and a person is being conducted. This conversation is recorded on the device. The word a (106) used in the sentence A (105) spoken by a human is searched (111) in the important word table (102), and if registered in this, this word and this word The data associated with the word table (10
Register (112) in 3). At the same time, the sentence A is also registered (115) in the sentence table (104).

Alternatively, when the word b (108) used in the sentence B (107) is searched (113) in the word table (103) and is already registered, it is determined that the frequency is high. Registers (114) the data attached to this word in the word table. At the same time, the sentence B is also registered (116) in the sentence table.

To summarize the above, it is essential to register important words, frequently used words, and sentences using them.

Next, a sentence (1 using the word b (110)
09) is registered in the sentence table, the sentence C
It is brought (117) and this device outputs it.

In other words, a past conversation sentence using a word which is a topic in the current conversation is brought and reproduced.

(Block Diagram) FIG. 2 is a block diagram of a personality forming conversation robot according to the present invention.

A CPU 201 is a central processing unit which controls various devices described later and performs calculations.

Reference numeral 202 DISP is a display device for displaying characters, and displays conversational sentences and the like.

Reference numeral 203SP is a speaker for emitting a sound.

Reference numeral 204KB is a keyboard, which is an operation panel of keys and switches for inputting characters to the apparatus and issuing operation instructions.

The 205MIC is a microphone that captures words spoken by humans.

206 ROM is a read-only memory for storing a program (PROG) and a character pattern (FONT) for operating this apparatus.

The 207 RAM is a random access memory having a work area for calculation and a temporary storage area for conversational sentences.

208HD is a hard disk in which various information is stored. Important word table (JTB
L), registered word table (TTBL), sentence table (B
TBL). These will be described later.

Further, there is a conversation sentence (KAIWA) for memorizing the conversation for one week.

Reference numeral 209BL is a bus line for connecting the above devices and exchanging information.

(Important Word Table) FIG. 3 shows the structure of the above-mentioned important word table. The composition is simply a list of important words. This is the word related to the ceremonial occasion related to life (301) and the word related to the incident (30).
2) is stored.

The important word table is used to determine whether the words used in the conversation are important or not.
Therefore, if the operator determines that it is important other than the above, the word (303) can be registered by himself.

(Registered Word Table) FIG. 4 shows the configuration of the registered word table described above.

The words registered here are important words used in conversation or words frequently appearing continuously used over a day.

The structure is as follows. First there is the word (401). Then there is the date this word was first used (402) and last used (403). Furthermore, the * mark (404) is a mark attached when the rank of the word is 4 or more. Next comes the rank (405). Ranks are from 1 to 6, with 6 being the highest.
The rank will be described later.

(Sentence Table) FIG. 5 shows the structure of the sentence table. The conversation sentences using the words registered in the above-mentioned registered word table are registered. The structure is such that the date of conversation (501) is first, and thereafter, it is composed of the speaker (502), the conversation sentence (503), and the registered word (504) used in the conversation sentence.

This sentence also disappears when all the registered words used in the sentence disappear.

(Rank of Registered Words) FIG. 6 is a rank table of registered words. There are 6 ranks from 1 to 6. Rank 1 words are words that have been used for two consecutive days. With the same idea, the word of rank 6 is a word used for 7 consecutive days, and is extremely frequently used. Also, the important word is rank 6.

(Diminishing Rank and Disappearing Registered Words) The rank of registered words decreases as the days go by. This is shown in FIG. That is, one week lowers one rank. Words with a rank of 3 or less initially disappear after a week after becoming rank 1. However, words with 4 ranks or higher initially fall to rank 1 and then maintain rank 1 as they are and do not disappear. This is because the frequently used words and important words are regarded as big topics and unforgettable words, and are not deleted. 4 (404) in FIG. 4 is necessary as a mark of this word that is not erased.

(Operation Processing) An example of operation processing of this apparatus is shown in FIG.
~ It demonstrates with reference to the flowchart of FIG. The program for this operation is stored in the ROM.

However, the flowcharts of FIGS. 8 to 11 are processes that operate during actual conversation execution, and the flowchart of FIG. 12 is a process of lowering the rank of registered words that operates once a week.

Now, the description of FIG. 8 will be made. First, the process of accepting human talk is performed. This may be processed by voice recognition from voice through the microphone, or may be accepted by the character code from the keyboard immediately (step S801).

The accepted conversational sentence is the KAI of HD as it is.
It is stored in the WA (step S802).

Words are extracted from this conversational sentence (step S803).

Whether or not the extracted word is an important word is judged by comparing it with the important word table (step S804).

As a result, if it is an important word, it is registered in the registered word table. Register the date, the symbol * that will not disappear, and the rank together. Rank 6 (step S80)
5).

Next, the sentence having this important word is registered in the sentence table. When registering, dates and words are also registered according to the format shown in FIG. 5 (step S806) By the above processing, it is determined whether or not the important words and sentences have been registered, and then all the word checks have been completed (step S807).

If not completed, the process returns to the check again.
When finished, the process proceeds to the flowchart of FIG.

Here, first, it is determined whether or not the checked word is registered 8 days or more ago. That is, it is checked whether or not the word has been registered in the past (step S1101).

If it has been registered in the past, the corresponding sentence is extracted and output in order to make use of it in the current conversation.
That is, the past conversational sentence is reproduced. The output may be converted into voice and output from the speaker, or the character code may be converted into a font and displayed on the display (step S1).
102).

On the other hand, if the word is not in the past, a new sentence is created and output. There are various methods for creating a new sentence, such as a method of inserting a word into a sentence as a skeleton having a plurality of patterns in advance, but existing techniques can be used, and details thereof will be omitted here (step S1103). .

When the above process is completed, the process returns to the conversation accepting process again.

On the other hand, if it is determined in step S804 that the word is not an important word, it will be described with reference to the flowchart of FIG.

First, it is determined whether or not the extracted word is already registered in the registered word table (step S901).

If registered, it is checked whether or not it was registered yesterday. That is, it is checked whether or not it has been continuously used (step S902).

As a result, if it is still used yesterday, the rank is raised. However, 6 is the highest rank. Also change the last usage date. If the rank is 4 or higher, an extinction prohibition mark * is also attached (step S903).

After that, the corresponding sentence is registered (step S904).

After that, in step S807, it is determined whether or not all the words have been checked.

On the other hand, if it has not been registered, or if it has already been registered, it was not registered yesterday, and if it is earlier, it proceeds to the flowchart of FIG.

Here, it is first determined whether or not the extracted word is used in the conversation yesterday. That is, if it has not been registered, but it has been used yesterday, it means that it has been used for two consecutive days, and will be registered for the first time. Since the conversational sentence is stored for at least one day, it is easy to check (step S1001).

If the word is used as a result of the check, this word is registered in the registered word table with rank 1. The use start / end date is also attached (step S1002).

After that, the sentence using the above word is also registered in the sentence table (step S1003).

After that, the process returns to step S807. If this word is not used even yesterday, it means that the word has not been used continuously and is not registered. Also in this case, the process returns to step S807.

Through the above processing steps, words and sentences can be registered and past sentences can be reproduced.

(Rank change / erasure processing) Next, the rank reduction or disappearance of registered words and sentence disappearance processing will be described. This process should be performed once a week. This is because the rank is lowered on a weekly basis. Figure 1
A description will be given according to the flowchart of No. 2.

First, looking at the registered word table, the rank of each word is lowered by one step (step S1201).

As a result of lowering the rank, it is determined whether or not the word disappears. As a matter of course, words with the erasure prohibition mark * do not disappear (step S1202).

If it disappears, the registered word is deleted from the registered word table (step S1203).

Further, the registered word is deleted from the registered word column of the sentence table of the corresponding sentence (step S1204).

As a result, it is determined whether or not the registered word column of the sentence table is empty (step S1205).

When it becomes empty, the sentence itself is deleted from the sentence table (step S1206).

Further, all the conversation records before that except the conversation record of the previous day are deleted (step S1207).

With the above processing, the rank of words is deleted, the words are deleted, the sentences are deleted, and the conversational sentences are deleted.

(Other Embodiments) In the above-described embodiments, the words that are frequently used among the registered words are required to be continuously used on a daily basis, but even if they are not continuous, they are ranked by the number of appearances in a certain number of days. You may attach it. Although the sentence to be registered is limited to the sentence including the registered word, the sentences before and after the sentence may be registered.

[0070]

As described above, in the present apparatus, the contents spoken by a human being are stored, past stories are reproduced according to the importance of the topic, and it is inserted in the present conversation, and it becomes natural. We realized a conversation as if we were talking to a human.

[Brief description of drawings]

FIG. 1 is a conceptual diagram of the operation of the device.

FIG. 2 is a block diagram of the device.

FIG. 3 is a configuration diagram of an important word table stored in a hard disk.

FIG. 4 is a configuration diagram of a registered word table stored in a hard disk.

FIG. 5 is a configuration diagram of a sentence table stored in a hard disk.

FIG. 6 is a rank reference table of registered words.

FIG. 7 is a reference table for rank reduction of registered words.

FIG. 8 is a flowchart of processing of this device.

FIG. 9 is a flowchart of processing of this device.

FIG. 10 is a flowchart of processing of the present apparatus.

FIG. 11 is a flowchart of the processing of this device.

FIG. 12 is a flowchart of the processing of this device.

Claims (3)

[Claims] 1. A conversational sentence input unit for inputting a conversational sentence, a conversational sentence storage unit for storing the conversational sentence, a conversational sentence output unit for outputting the conversational sentence, and a conversational sentence ranking unit for ranking the conversational sentence. The present invention is characterized by further comprising a conversation sentence deletion determining means for determining whether or not the conversation sentences ranked by the conversation sentence ranking means are to be erased, and a conversation sentence deleting means for deleting the conversation sentences by the conversation sentence deletion determining means. Data storage type conversation device. 2. The data storage type conversation apparatus according to claim 1, wherein the importance of the words used in the conversation sentence or the frequency of use is used as the basis for the discrimination by the conversation sentence erasure discrimination means. 3. The data storage type conversational apparatus according to claim 2, wherein the output timing of the conversational sentence output means depends on the type of the input word by the conversational sentence input means.