JP2002169591A - Simulated conversation system and information storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily change a conversation according to a user. SOLUTION: A user information acquisition part 234 obtains the profile of a user and generates user data 516. A reply determination part 232 determines a simple response word and a next question for an adult or child according to the profile of the user. Then, a question for starting the conversation is outputted through a speaker 30 and when a voice to the outputted question is inputted, a voice recognition part 210 recognizes the voice (input word) (determines a registered word corresponding to the voice). The reply determination part 232 determines a simple response word and a next question according to the registered word in a determined simple reaction word group and a next question group and a voice synthesis part 220 synthesizes the voices of the determined simple response word and next question and outputs them through the speaker 30 to carry on the conversation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a simulated conversation system for outputting a response to a word input by a user, and an information storage medium.

[0002]

2. Description of the Related Art Conventionally, there has been known a simulated conversation system which outputs a response to a voice input by a user and has a conversation with the user. In such a simulated conversation system, for example, when the user is a child and when the user is an adult, the words frequently used by the user are different, and the contents of replies to the words are different. Therefore, in the conventional simulated conversation system, a cartridge or a CD-ROM in which different words (registered words for recognizing a user's voice, replies to the registered words, etc.) are registered according to the profile (age, gender, etc.) of the user. A plurality of such devices are prepared, and a voice corresponding to the user's profile is realized by recognizing the user's voice based on the words registered in each cartridge or CD-ROM and outputting a response. .

[0003]

However, the conventional simulated conversation system has a problem that a plurality of cartridges must be prepared. In addition, there is a problem that the cartridge has to be changed every time the user's profile changes, which is troublesome.

[0004] In the conventional simulated conversation system, since a reply is set for each registered word for one question,
If you input the same input words to the same question, the same answer will be output, and users who have many conversations will be able to read ahead, become less interesting, and get bored easily. there were.

[0005] An object of the present invention is to realize a simulated conversation system that can easily change the conversation according to the user.

[0006]

A simulated conversation system according to the present invention (for example, a conversational toy 1 shown in FIG. 1).
Stores a plurality of registered words and a plurality of replies corresponding to one registered word (for example, the storage unit 500 shown in FIG. 2), and stores the registered words included in the words input by the user. A recognizing means (for example, a voice recognition unit 210 shown in FIG. 2) for recognizing from the registered words stored in the storage means, and a reply corresponding to the registered word recognized by the recognizing means. Determining means (for example, a response determining unit 234 shown in FIG. 2) for determining a response according to the condition from among the responses stored in the storage means, and an output means for outputting the response determined by the determining means ( For example, a speaker 30) shown in FIG. 2 is provided, and a simulated conversation with the user is performed by repeatedly executing a series of processes of inputting a word, determining a response, and outputting by the user. Even the same words input by chromatography, replies the determining means determines is characterized in that not always the same.

According to a ninth aspect of the present invention, computer-executable software for performing a simulated conversation with a user by repeatedly executing a series of processes of inputting a word, determining a response, and outputting the word is stored. The information storage medium (for example, the storage unit 500 shown in FIG. 2) includes a plurality of registered words, a plurality of replies corresponding to one registered word, and a registered word included in the word input by the user. Recognition information for recognizing from the plurality of registered words, and determination information for determining a response according to a given condition from among a plurality of responses corresponding to the recognized registered words, And output information for outputting the reply, and even if the words input by the user are the same,
It is characterized by storing information for preventing the output response from being the same.

Here, the given conditions include, for example, the profile of the user (age (may be age, adult or child), gender, number of conversations, level (advanced, elementary) Etc.), the profile of the virtual character (age, gender, etc.), the situation of conversation (favorable conversation, conversation time, the number of questions and answers exchanges, the percentage of unrecognized words in the input words, etc.) No. Also,
A combination of these conditions may be used.

According to the first or ninth aspect of the present invention, even if the word input by the user for one question is the same, the output response is not necessarily the same, so that the conversation can be more effectively performed. It can be various. Therefore, even if the user has spoken many times, the output response is difficult to predict, which makes the user more interesting and hard to get tired of.

[0010] In the simulated conversation system according to the first aspect of the present invention, the reply includes a question, and the simulated conversation is a simulated conversation in a question-and-answer format. Is also good.

[0011] In the information storage medium according to the ninth aspect of the present invention, the reply includes a question, and the simulated conversation is a simulated conversation in a question-and-answer format. Is also good.

According to the second or tenth aspect of the invention, since the conversation is in a question-and-answer format, the topic in the conversation can be easily changed by changing the question included in the reply. Can be changed in various ways.

According to a third aspect of the present invention, in the simulated conversation system according to the second aspect of the present invention, the storage means stores the plurality of registered words in a registered word group classified based on a question. The recognition means selects the registered word group corresponding to the question included in the answer previously output by the output means from the registered word group stored in the storage means, and stores the registered word group. From among them, the registered words included in the words input by the user may be recognized.

[0014] In the information storage medium according to the tenth aspect of the present invention, as in the eleventh aspect, the plurality of registered words are stored by being divided into registered word groups classified based on a question. The recognition information is information for selecting a registered word group corresponding to the question included in the previously output response and recognizing a registered word included in the word input by the user from the registered word group. May be included.

According to the third or eleventh aspect of the present invention, when recognizing a registered word included in a word input by a user, only a registered word group corresponding to a question included in a previously output response is recognized. Since the registered word can be recognized with respect to, the time required for recognizing the input word can be reduced. Further, for example, by dividing one registered word into a plurality of registered word groups and dividing, one registered word can correspond to a plurality of questions.

According to a fourth aspect of the present invention, in the simulated conversation system according to any one of the first to third aspects, the judging means for judging the evaluation value of the conversation situation in the current simulated conversation (for example, , The response determination unit 23 shown in FIG.
2), wherein the storage means includes, as registered words, an affirmative registered word meaning a positive response (for example, a registered word 3-1 (yes) in the conversation data 520c shown in FIG. 10) and a negative response Negative registered word (eg, registered word 3-2 (No) in conversation data 520c shown in FIG. 10)
And a polysemy registered word (for example, a registered word 3-3 (yes) in the conversation data 520c shown in FIG. 10) meaning an ambiguous response that can be either affirmative or negative, and stored. The determining means, when the recognizing means recognizes a registered word included in a word input by a user as the ambiguous registered word, according to the evaluation value determined by the determining means, A word may be regarded as the positive registered word or the negative registered word, and a response corresponding to the regarded positive registered word or negative registered word may be determined.

According to a twelfth aspect of the present invention, in the information storage medium according to any one of the first to third aspects, determination information for determining an evaluation value of a conversation situation in a current simulated conversation is provided. As the registered words, a positive registered word meaning a positive response, a negative registered word meaning a negative response, and an ambiguous registration meaning an ambiguous response that can be either positive or negative. Word, and the decision information is, when the registered word included in the word input by the user with the recognition information is recognized as the ambiguous registered word, the evaluation value determined by the determination information Accordingly, the ambiguous registered word may be regarded as the positive registered word or the negative registered word, and information for determining a response corresponding to the regarded positive registered word or negative registered word may be included.

According to the fourth or twelfth aspect of the present invention, even if the word input by the user is the same, the interpretation of the meaning of the word is changed according to the conversation situation.
The response can be changed according to the interpretation. That is,
Conversations can be more diverse. for that reason,
Even when the user has repeated conversations, the output response is difficult to predict and can be made more difficult to get tired of.

In the simulated conversation system according to any one of the first to fourth aspects of the present invention, the response to be output may be determined based on the profile information of the user. good.

According to a thirteenth aspect of the present invention, in the information storage medium according to any one of the ninth to twelfth aspects, the information is output based on the profile information of the user and the profile information of the user. Information for determining a response may be stored.

According to the fifth or thirteenth aspect of the present invention, a reply according to the user's profile can be output, and conversation suitable for the user can be easily realized.

In the simulated conversation system according to the fifth aspect of the present invention, as in the sixth aspect of the present invention, the reply is a virtual character (for example, the robot 2 shown in FIG. 1) whose profile is virtually set. ) Is a response set as a word issued by the user, and the determining means may determine a response to be output based on a difference between the profile information of the user and the profile of the virtual character.

In the information storage medium according to the thirteenth aspect of the present invention, as in the fourteenth aspect, the reply is a virtual character (for example, the robot 2 shown in FIG. 1) whose profile is virtually set. ) Is a response set as a word issued, and the determination information may include information for determining a response to be output based on a difference between the profile information of the user and the profile of the virtual character.

Here, the user profile information and
The differences in the profile of the virtual character include, for example, differences in age, gender, and the like.

According to the sixth or fourteenth aspect of the present invention, it is possible to output a response according to the difference between the user's profile and the virtual character's profile. Even if is input, if the character performing the conversation is different, the same response is not always output, so the conversation can be made more diverse, making the output response difficult to predict, The user can be bored, and the user can enjoy the conversation many times.

According to a seventh aspect of the present invention, in the simulated conversation system according to the fifth aspect, the storage means stores the plurality of replies in a group of replies corresponding to the assumed profile information of the user. The determination means may select a response group corresponding to the user's profile information from the response group stored in the storage means, and determine a response corresponding to the registered word from the selected response group. good.

In the information storage medium according to the present invention, the plurality of replies are stored by being classified into a group of replies corresponding to the assumed user profile information. The determination information may include information for determining a response corresponding to the registered word from a group of responses corresponding to the profile information of the user.

According to the seventh or fifteenth aspect of the present invention, since a reply according to the profile of the user can be easily determined, a conversation suitable for the user can be easily realized.

In the simulated conversation system according to any one of the first to seventh aspects, the simulated conversation is a simulated conversation by voice, and the recognition means includes: The registered words included in the voice data input by the user are recognized from the registered words stored in the storage means, the storage means stores voice data to be output as the reply, and the output means, Reading voice data corresponding to the reply from the storage means,
It may be output as audio.

In the information storage medium according to any one of the ninth to fifteenth aspects, as in the sixteenth aspect, the simulated conversation is a simulated conversation by voice and is output as the reply. Voice data, the recognition information includes information for recognizing a registered word included in the voice data input by the user from the plurality of registered words, and the output unit corresponds to the reply. It may include information for reading out audio data and outputting it as audio.

According to the invention described in claim 8 or claim 16, since conversation can be performed by voice, even a user who cannot read and write characters or cannot input characters using a keyboard can easily talk. Can be performed. Further, since the conversation is performed by voice, the user can enjoy the work while performing work such as driving a car.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is an external view showing an example in which the present invention is applied to a conversational toy 1. As shown in FIG. 1, the conversational toy 1 includes a robot 2 and a pedestal 3 on which the robot 2 is placed. The microphone 4 detects a voice spoken by the user to the robot 2, and the detected user's A response to the voice is output from the speaker 6. The user listens to a response from the robot 2 to the voice spoken to the robot 2 and enjoys a conversation with the robot 2.

FIG. 2 is a block diagram showing an example of functional blocks according to the present embodiment. As shown in the figure,
The functional block according to the present embodiment includes a voice input unit 10, a processing unit 200, a speaker 30, and a storage unit 500.

The voice input unit 10 corresponds to the microphone 4 shown in FIG. 1, and outputs the input voice to the processing unit 200.

The processing section 200 recognizes the voice, determines a response to the recognized word, and performs processes such as a process of synthesizing the voice of the response. The processing unit 200 includes a voice recognition unit 210, a voice synthesis unit 220, a response determination unit 232, a user information acquisition unit 234, a history generation unit 236, and a clock 240.

The voice recognition unit 210 compares the voice (input word) input from the voice input unit 10 with the voice data of the registered word registered in the voice recognition dictionary 530, and determines the input voice. To determine if it is a registered word. At this time, the speech recognition unit 210 transmits the conversation data 52 described later.
At 0, it is determined which of the registered words registered in response to the question output immediately before that. That is, the input voice is compared with the voice data of the registered word corresponding to the question output immediately before in the conversation data 520 among the voice data registered in the voice recognition dictionary 530, and Is recognized as one of the registered words corresponding to the question output to. Also, the voice recognition unit 21
0 determines that the input word is an unrecognized word if the input word is not a registered word for the question output immediately before or if the input word cannot be recognized.

The comparison (recognition) of the voice data is performed as follows.
It is realized by using a word spotting technique which is a conventional technique. The word spotting technique is a technique for detecting, by pattern matching, whether or not waveform data of a short voice to be detected is included in waveform data of a long voice. According to this technique, it is possible to determine whether or not a long speech including a registered word input by a user includes a registered word, that is, whether or not the input is a registered word.

FIG. 3 is a diagram showing conversation data 520a which is an example of a data structure for one question among the conversation data 520. As shown in the figure, a registered word is set for a question, and a favorable point corresponding to each registered word is set. Then, for each registered word, a child's hammer and the next question, and an adult's hammer and the next question are set.

In FIG. 3, for example, the question is question 1-
1 (what do you like food?), In this case, registered words 1-1 (chocolate), registered words 1-2 (carrots), registered words 1-3 (baked meat),
Words indicating various foods such as... Are registered (registered words having an attribute of “food” are set). Then, a favorable point is set for each registered word. For example, it is set in advance according to the character or the like assumed for the character of the robot 2. In FIG. 3, for example, a high liking point is set for sweet foods, and a low liking point is set for vegetables.

For example, for a registered word 1-1, a child hammer b1 (I also like, but worry about tooth decay) is set, and as an adult hammer, a hammer a1 (I like, But it's fat). Also, registered word 1
As for the next question for children, question 1-2 (has you brushed your teeth properly) is set for the next question for children, and for the next question for adults, question 1-5 (who is fat?) ?) Is set.

The voice recognition dictionary 530 is a dictionary that associates registered words with their voice data. For example, since a registered word is set in the conversation data 520 in accordance with the question, for example, as a registered word for a question such as “What do you like about food?”, A registered word whose attribute is “food” is set. And the registered word includes "chocolate". In addition, a registered word whose attribute is “food” is set as a registered word for a question such as “What is the disliked food?”, And the registered word includes “chocolate”. As described above, in the conversation data 520, the same registered word is registered for a plurality of questions, but the speech recognition dictionary 530 is used.
Stores only one voice data for each registered word. Thus, the registered words set in the conversation data 520 are the text data of the registered words corresponding to the voice data,
Or, it is replaced with the identification number of the registered word.

Also, for example, "Tall" (height) as a registered word for a question such as "Is that building 80 stories high? What do you think?" Or "This clothes is 100,000 yen. Even if the meaning is different, such as “expensive” (price) as a registered word for a question such as “?”, Since the voice is the same “takaine”, one voice is stored in the voice recognition dictionary 530. You only need to register the data.

When starting a conversation, the user information acquisition unit 234 acquires a user profile (age, gender, level (beginner / advanced), etc.) and acquires user data 516.
Is performed. As a method for acquiring the user's profile, for example, at the start of a conversation, the user may be asked for the profile of the user by conversation, and profile information obtained by voice input by the user may be acquired. It may be obtained by input or the like.

FIG. 4 is a diagram showing an example of the data structure of the user data 516. As shown in the figure, age, gender, and level are set as user data.
In FIG. 4, for example, "10" is set as the age, "male" as the gender, and "elementary" as the level.

It is to be noted that not age, for example, teens, 20
It may be of the age of a teenager, and may be an adult or a child.

In addition to the age, gender and level, for example, the number of conversations may be obtained as a profile.

The response determination unit 232 determines whether the conversation is for adults or children with reference to the user data 516. For example, in the case of the user data 516 as shown in FIG. 4, since the age is 10, a conversation for a child is determined.

The answer determining unit 232 refers to the conversation data 520 to determine the answer and the question for the input word that is input for the question and recognized by the voice recognition unit 210.
At that time, if the conversation was determined to be for children, determine the child's hammer and the next question, and if the conversation was determined to be for adults, ask the adult's hammer and the next question. decide.

For example, the conversation data 520a shown in FIG.
If the input word for the question 1-1 in the above corresponds to the registered word 1-1 and the user is a child (for example, the age is “15” or less) from the age set in the user data 516, the answer determination unit 232 determines the companion b1 for the child as the companion to the registered word 1-1, and determines the question 1-2 as the next question. If the user is an adult (for example, the age is “16” or more), the companion a1 for the adult is determined as the companion to the registered word 1-1, and the question 1-5 is determined as the next question.

In the next question, if a connective is set before the question, as in the case of connective + question 1-3 in the next question, the response determiner 232 sets the plurality of connectives set in the connective data 518. One of the conjunctions is randomly determined. The conjunction may be determined according to the next question. In the present specification, the conjunctions are, for example, "by the way", "by the way", "it looks like",
Refers to words spoken when changing topics in conversations such as "yes", "later", ...

As described above, since the hammer and the next question are different for adults and children, for example, if the user is a child, make sure that conversations such as driving and drinking / cigarettes are inappropriate for the child. The next question or the like can be set, and a conversation according to the user's profile (age) can be realized.

In the conversation data 520a, it is assumed that the child and adult socks and the next question are set, but, for example, only the soujin is prepared for adults and children and the next question is one. There may be one, and there is only one companion, and only the next question may be divided for adults and children.

In addition, two types of hammering for adults and children and the following questions are set. For example, for each age such as 10s, 20s, 30s, etc., or For elementary school students, junior high school students, high school students, university students, adults, it is also possible to set a hammer and the next question, and further, for example, a gender hammer and the next question for men, women, etc. It may be set in advance.

Further, for example, it is also possible to set the liking point, the hammer, and the next question for each level (for example, for the elementary level and the advanced level). For example, conversation data 520
In a, for the elementary level, a high liking point (“+1”) is set for all registered sweet words, and a high liking (for example, (I also like)) is set. You. On the other hand, in the case of advanced use, a high likability point is set only for the registered word 1-1 (chocolate) and a high liking hammer is set. For a registered word of a sweet thing, a normal favorable point (“0”) is set, and a normal word (for example,
A companion is set.

That is, in an advanced conversation, it is desired to input a registered word more accurately with respect to the character assumed for the character of the robot 2. Further, for example, in the case of a conversation situation in which the accumulated favorable points accumulated in the cumulative favorable point data 512 described later are high, the conversation is continued,
In the case of a conversation situation where the accumulated favorable liking point is low, if the conversation is ended, the continuity of the conversation can be changed between the elementary level and the advanced level.

As described above, depending on the level, the hitting for the same registered word, the ease with which the conversation can be continued, and the like change, so that even a user who has spoken many times can be made less tired. .

Further, for example, a plurality of characters of the robot 2 (characters given a given character) for talking with the user are set, and for each character, the favorable points for the registered words, the so-called hammer and the next question May be set. In other words, even if the same registered word is input for the same question, the external shape of the robot 2 is the same, but the liking point, the companion, and the next question change according to the character to be spoken. It will be.

Further, age, gender, etc. are set as profiles assumed for each character, older and younger companions and the following questions are set for each character, and the age of the character performing the conversation is set. And the user's age, and if the user is older than the character, determine the older hammer and the next hammer and the next question corresponding to the registered word entered from the question. If the user is younger, it may be possible to determine a younger partner and a next agent based on the registered word input from the next question.

As described above, even when the same registered word is input for one question, the output of the hammer and the next question are changed according to the user's profile and character settings, and the conversation is performed. Changes. For this reason, conversations can be made more diverse, and even if the user has repeated conversations, different conversations can be enjoyed each time.

FIG. 5 shows the accumulated favorable liking point data 512.
FIG. 3 is a diagram showing an example of the data configuration of FIG. The accumulated favorable positivity point data 512 stores a value obtained by adding the favorable positivity points of the registered words determined for the input words input up to now. The accumulated favorable liking point data 51
2 is updated by the history generation unit 236 described later.

FIG. 6 is a diagram showing an example of the data configuration of the conversation status data 514. As shown in the drawing, the conversation situation data 514 is set with the conversation situation corresponding to the value of the accumulated favorable liking point.

For example, as shown in FIG. 5, when the cumulative favorable points stored in the cumulative favorable points data 512 is +3, the conversation status data 514 shown in FIG.
In the above, since "B" is set as the conversation status for the accumulated favorable liking point +3, the answer determination unit 232
Determines that the conversation status is “B”.

The history generation unit 236 updates the above-described cumulative favorable liking point data 512 based on the favorable liking points of the registered word determined for the input word.

The history generation unit 236 generates not only the accumulated favorable liking point data 512 but also the total conversation time, the number of conversations, the unrecognized word (the recognized input word) among the previously input words. Is not a registered word, or an input word when an input word cannot be recognized is referred to as an unrecognized word.). Further, the conversation status data 514 may set the conversation status in accordance with any one or more of the accumulated favorable liking points, the total conversation time, the number of conversations exchanged, the ratio of unrecognized words in the input words, and the like. good. Note that the total conversation time is measured based on a clock signal input from the clock 240.

The voice synthesizing unit 220 synthesizes the voice data of the hammer and the question (including the conjunction) determined by the response determining unit 232 and performs D / A conversion to output the voice from the speaker 30. . Note that the voice data of the hammer and the question may be stored in the storage unit 500 or may be held in the voice synthesis unit 220.

The function of the processing unit 200 described above is
And RISC type CPU, DSP, and other hardware.

The speaker 30 outputs the voice synthesized by the voice synthesizer 220. The speaker 30 corresponds to the speaker 6 shown in FIG.

The storage unit 500 stores a conversation program 510,
The voice recognition dictionary 530 and the conversation status data 51 described above.
4. Accumulated favorable point data 512, user data 516, conjunction data 518, and conversation data 520 are stored. The function of the storage unit 500 is a CD-ROM,
It can be realized by hardware such as an IC card, MO, FD, DVD, hard disk, and memory. As described above, the processing unit 200 performs various processes based on the programs, data, and the like stored in the storage unit 500.

Next, the operation relating to the conversation processing in the present embodiment will be described with reference to the flowchart shown in FIG.

First, the user information acquisition section 234 acquires a user profile and generates user data 516 (step S1). Then, the hammer group / next question group is determined according to the user's profile (for example,
Decide whether it is for adults or children. (Step S2). Next, the response determination unit 232 outputs a question for starting conversation through the speaker 30 (step S).
3). When a voice corresponding to the output question is input, voice recognition unit 210 recognizes the voice (determines a registered word corresponding to the voice (input word)) (step S).
4).

Then, the response determining unit 232 determines in step S3
The hammer and the next question are determined according to the registered word from the hammer group / next question group determined in (Step S5),
The voice synthesis unit 220 synthesizes the voices of the determined hammer and the next question, outputs the voice by the speaker 30 (step S6), and returns to step S4. And step S
The voice input for the question output in step 6 is recognized, and the subsequent processing is repeated to continue the conversation.

Next, an example of a hardware configuration capable of realizing the present embodiment will be described with reference to FIG. In the device shown in FIG.
AM1004, information storage medium 1006, sound generation IC10
08, voice recognition IC 1012, I / O port 1014,
Clocks 1026 are connected by a system bus 1016 so that data can be input and output to and from each other. And sound generation IC1
008, a speaker 1018 is connected, and a voice recognition IC
A microphone 1020 is connected to 1012, and a communication device 1024 is connected to the I / O port 1014.

The information storage medium 1006 mainly stores programs, sound data, play data, and the like.
It is composed of a semiconductor memory or an optical or magnetic recording medium. This information storage medium 1006 corresponds to the storage section 500 in FIG.

The microphone 1020 corresponds to the voice input unit 10 in FIG. 2, and detects a voice uttered by the user. The voice recognition IC 1012 is an integrated circuit for recognizing the voice detected by the microphone 1020. Specifically, the voice recognition IC 1012 includes a recognition engine having a continuous voice recognition function and a word spotting function known as a conventional technology, a recognition dictionary used for the recognition engine, and the like, and converts analog voice data into digital data. Is converted into text data. Therefore, the registered word in FIG. 3 may be a proper noun itself or a part of a sentence depending on the function of word spotting.

According to the program stored in the information storage medium 1006, the system program stored in the ROM 1002, the voice input from the microphone 1020, etc.
The PU 1000 controls the entire apparatus and performs various data processing. The RAM 1004 is a storage unit used as a work area or the like of the CPU 1000.
6 or given contents of ROM 1002 or CPU 100
The operation result of 0 is stored. The storage unit 50 shown in FIG.
Of the data stored in 0, the accumulated favorable liking point data 512 may be stored in the RAM 1002.

The sound generation IC 1008 is the information storage medium 100
6 is an integrated circuit that generates sound based on information stored in the ROM 1002.
018.

The communication device 1024 exchanges various kinds of information used inside the device with the outside. The communication device 1024 is connected to another device to send and receive given information according to a conversation program, etc. It is used for transmitting and receiving information such as a conversation program and data via the Internet.

The clock 1026 is a clock circuit for measuring the current time and outputting a clock signal to the CPU 1000 as needed.

The various processes described with reference to FIGS. 1 to 6 include an information storage medium 1006 storing a program for performing the processes shown in the flowchart of FIG. 7, a CPU 1000 operating according to the program, and a sound generation device. IC1
008, the voice recognition IC 1012 and the like.
The processing performed by the voice recognition IC 1012 and the like is performed by the CPU.
It may be performed by software using a 1000 or a general-purpose DSP.

As described above, according to the present invention, even if the words input by the user for one question are the same, the output hammer and the next question change according to the user data 516 and the conversation data 520. Therefore, the conversation can be made more diverse, and even if the user has conducted the conversation many times, it is difficult to predict the response to be output, and the interest can be further increased.

The present invention is not limited to those described in the above embodiment, but can be variously modified. For example, in the above-described embodiment, it is assumed that the conversation data 520 includes a hammer and the next question corresponding to the user data 516.
At 20, a hammer and the next question according to the conversation situation may be set. In that case, the response determination unit 232 refers to the accumulated favorable liking point data 512 and the conversation situation data 514 to determine the current conversation situation, and decides a companion and a next question according to the conversation situation. Becomes

FIG. 9 is a diagram showing an example of the data configuration of conversation data 520b in which a companion and the next question are set for each conversation situation. FIG. 9 shows a companion for the child and the next question, and the illustration for the adult is omitted, but the companion and the next question are set for each conversation situation as in the case of the child. Become. In the conversation data 520b, a plurality of registered words are set for a question, and a favorable impression point is set for each registered word. Further, for each registered word, a hammer and the next question are set for each conversation situation for children and adults.

For example, FIG. 9 shows a question 2-1 (sea lover?).
Data 520b with respect to the question 2-1.
Words 2-1 (like), registered words 2 as registered words for
-2 (dislike), registered words 2-3 (OK), ..., etc. are set. And the conversation situation is "A"
In the case of "B", in the case of "B", in the case of "C", the hammer and the next question are set. For example, with respect to the registered word 2-2, if the favorable impression point is "-1" and the conversation situation is "A", the companion is c2 (sorry), and the next question is the question 2-2. (Let's go together), but even if the same registered word 2-2 is entered,
If the conversation situation is "C", the souetsu will output a souvenir e2 (I'm not happy), no next question, and an end message (Goodbye).

Further, for example, the conversation data may have a data structure such as conversation data 520c shown in FIG. FIG. 10 shows conversation data 5 in the case where words that can be interpreted in a plurality of meanings (ambiguous) are registered as registered words.
It is a figure showing an example of the data composition of 20c.

As shown in FIG. 10, registered words 3-1 (yes), registered words 3-2 (yada), registered words 3-
When 3 (good),... Are set, for example, the registered word 3-3 (good) can be interpreted as an affirmative meaning (the same meaning as the registered word 3-1), or a negative. (The same meaning as the registered word 3-2). When such ambiguous words are included in the registered words, they may be interpreted according to the conversation situation.

For example, in FIG. 10, when the conversation status is “A” or “B” for the registered word 3-3,
When the similarity point, the companion, and the next question similar to the registered word 3-1 are set, and the conversation status is “C”,
The same liking point, companion, and next question as the registered word 3-2 are set. That is, when the conversation status is "A" or "B", the registered word 3-3 is regarded as having a positive meaning (registered word 3-1), and when the conversation status is "C", the registered word is registered. Considering the word 3-3 as a negative meaning (registered word 3-2), the favorability point, the companion, and the next question will be determined. That is, even if the words input by the user are the same, the interpretation of the meaning of the words can be changed according to the conversation situation.

It is to be noted that the interpretation of the meaning of words may be changed at random, for example, not according to the conversation situation.

Further, for example, in the above-described embodiment, the case where the present invention is applied to the interactive toy 1 has been described. However, the present invention can be applied to, for example, a personal computer or an arcade game device. FIG. 11 is an external view showing an example of the arcade game device. In FIG.
A display 18, a microphone 14, and a speaker 16 are provided. The player inputs voice to the microphone 14, listens to voices such as a question and a match output from the speaker 16, and enjoys a conversation with the character displayed on the display 18.

When the present invention is applied to an apparatus having a display, not only voice conversation but also text conversation may be performed. That is, the user may input character data, and display a hammer, a question, etc. corresponding to the word on the display in characters. Further, the words input by the user may be input by voice, and the response may be output by characters. Conversely, the words input by the user may be input in characters, and the response may be output in voice.

Further, for example, the simulated conversation system of the present invention can be incorporated into an electric appliance such as a telephone or a remote control device, or can be attached to a wall or the like of a room as a device so that a user can talk with an object or a room. You can feel like you are doing That is,
The greatest feature of the simulated conversation system of the present invention is that a continuous conversation can be executed without giving a mechanical feeling, and the conversation can be enjoyed. Therefore, the application target of the present invention may be a doll or an arcade game device, but may be applied to an article far from conversation, such as a refrigerator or a telephone. Alternatively, the chair or the wall may be realized as a simulated conversation system by being configured in a panel shape and installed on a chair or a wall. In this case, by providing the simulated conversation system with the data of the contact and the question when the chair or the wall is personified, it is possible to make it feel as if the user is talking with the chair or the wall.

It should be noted that the expression and movement of the robot 2 described in the above embodiment, the expression of the character and the movement displayed on the display of the arcade game device shown in FIG. Alternatively, it may be changed according to the favorable points of the registered word corresponding to the input word.

Further, for example, the present invention can be applied to a conversation other than the case of continuing in a question-and-answer format. For example, when the present simulated conversation system is applied to a home air conditioner, when the user inputs a voice "cooling ON", the user turns on the cooling switch and, for example, according to the age and gender of the user. Change the response, such as replying, "If it's too cold, it's not good for your body" or "Replying to play outside, not just in the room". In this case, since the user is limited to the family, a profile corresponding to the number of the family is set in the user data 516, and more detailed profiles such as illness status as well as age and gender are stored. The conversation data 520 includes, for each family member,
A companion based on the user data 516 may be set.

Further, in the present embodiment, Japanese conversation has been described as an example. However, the present invention is not limited to Japanese, but may be a language of another country or a dialect.

[0095]

According to the present invention, even if the word input by the user for one question is the same, for example, the interpretation of the meaning of the word is changed according to the conversation situation, and a reply is made according to the interpretation. Is changed, or the response is changed according to the profile of the user or the profile of the virtual character. Therefore, the output response is not always the same.
Therefore, conversation can be made more diverse. Therefore, even if the user has spoken many times, the output response is difficult to predict, which makes the user more interesting and hard to get tired of.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an example of a conversational toy according to the present embodiment.

FIG. 2 is a block diagram illustrating an example of a functional block according to the present embodiment.

FIG. 3 is a diagram illustrating an example of a data configuration of conversation data.

FIG. 4 is a diagram illustrating an example of a data configuration of user data.

FIG. 5 is a diagram illustrating an example of a data configuration of accumulated favorable liking point data.

FIG. 6 is a diagram illustrating an example of a data configuration of conversation status data.

FIG. 7 is a flowchart showing an operation related to conversation processing in the present embodiment.

FIG. 8 is a diagram illustrating an example of a hardware configuration that can implement the present embodiment;

FIG. 9 is a diagram illustrating an example of a data configuration of conversation data.

FIG. 10 is a diagram illustrating an example of a data configuration of conversation data.

FIG. 11 is a diagram illustrating a case where the present invention is applied to an arcade game device.

[Explanation of symbols]

 Reference Signs List 10 input unit 200 processing unit 210 voice recognition unit 220 voice synthesis unit 232 reply determination unit 234 user information acquisition unit 236 history generation unit 240 clock 30 speaker 500 storage unit 510 conversation program 512 accumulated favorable liking point data 514 conversation status data 516 user data 518 connective data 520 conversation data 530 dictionary for speech recognition

Continuing on the front page (72) Inventor Yasunori Tashiro 2-8-5 Tamagawa, Ota-ku, Tokyo Namco Corporation (72) Inventor Takashi Yamazaki 2-8-5 Tamagawa, Ota-ku, Tokyo Namco Corporation (72 ) Inventor Satoshi Yamamoto 2-8-5 Tamagawa, Ota-ku, Tokyo Tokyo, Japan (72) Inventor Yoshihiro Nagasaki 2-8-5 Tamagawa, Ota-ku, Tokyo Namco Corporation (72) Inventor Mitsuru Ogata Tokyo 2-8-5 Tamagawa, Ota-ku, Tokyo Namco Corporation (72) Inventor Masahide Uchida 2-8-5 Tamagawa, Ota-ku, Tokyo F-term (reference) 2C001 AA11 BA07 BB05 BB06 CA07 CB01 CB03 CC01 CC08 2C150 AA05 BA17 CA01 DF03 DF33 ED42 EE02 EF30 5B091 AA15 CB12 CB32 CC01 CC15 5D015 LL10

Claims (16)

[Claims] 1. A storage means for storing a plurality of registered words and a plurality of replies corresponding to one registered word, and a registered word included in a word inputted by a user is stored in said storage means. A recognizing means for recognizing from the registered words, and a response corresponding to a given condition, which is a response corresponding to the registered word recognized by the recognizing means, is determined from the responses stored in the storage means. A simulating conversation with the user by repeatedly executing a series of processes of inputting a word, determining a reply, and outputting the word by the user, comprising: In addition, a simulated conversation system is characterized in that even if the words input by the user are the same, the answer determined by the determining means is not necessarily the same. 2. The simulated conversation system according to claim 1, wherein the reply includes a question, and the simulated conversation is a simulated conversation in a question-and-answer format. 3. The storage unit according to claim 2, wherein the storage unit stores the plurality of registered words in a group of registered words classified based on a question, and the recognizing unit outputs the plurality of registered words by a previous output by the output unit. A registered word group corresponding to the question included in the received answer is selected from the registered word group stored in the storage unit, and a registered word included in the word input by the user is recognized from the registered word group. A simulated conversation system characterized by the following. 4. The method according to claim 1, further comprising: a determination unit configured to determine an evaluation value of a conversation situation in a current simulated conversation, wherein the storage unit includes a registration word indicating a positive response as a registration word. Word, a negative registered word meaning a negative response, and a polysemy registered word meaning a polysemy response that can be taken either positively or negatively, and stored. When the registered word included in the word input by the user is recognized as the ambiguous registered word, the ambiguous registered word is recognized as the positive registered word or the negative according to the evaluation value determined by the determining unit. A simulated conversation system characterized by considering a registered word and determining a reply corresponding to the recognized positive registered word or negative registered word. 5. The simulated conversation system according to claim 1, wherein a response to be output is determined based on profile information of the user. 6. The response according to claim 5, wherein the response is a response set as a word uttered by a virtual character whose profile is virtually set, and wherein the deciding unit determines the profile information of the user and the profile of the virtual character. A simulated conversation system, wherein a response to be output is determined based on the difference between the two. 7. The method according to claim 5, wherein the storage unit classifies and stores the plurality of replies into a group of replies corresponding to the assumed profile information of the user. A simulated conversation system, wherein a corresponding response group is selected from the response groups stored in the storage means, and a response corresponding to the registered word is determined from the selected response group. 8. The simulated conversation according to claim 1, wherein the simulated conversation is a simulated conversation by voice, and the recognizing unit stores a registered word included in voice data input by a user in the storage unit. The storage means stores voice data to be output as the reply, and the output means reads out the voice data corresponding to the reply from the storage means and outputs it as voice. A simulated conversation system characterized by the following. 9. An information storage medium storing computer-executable software for performing a simulated conversation with a user by repeatedly executing a series of processes of inputting a word, determining a response, and outputting the word by the user. A plurality of registered words, a plurality of replies corresponding to one registered word, recognition information for recognizing a registered word included in a word input by a user from among the plurality of registered words, Input from the user, including: determination information for determining a response according to a given condition from among a plurality of responses corresponding to the registered word, and output information for outputting the determined response. An information storage medium for storing information for ensuring that an output response is not necessarily the same even if the input word is the same. 10. The information storage medium according to claim 9, wherein the reply includes a question, and the simulated conversation is a simulated conversation in a question-and-answer format. 11. The registered word according to claim 10, wherein the plurality of registered words are stored in a group of registered words classified based on a question, and the recognition information corresponds to a question included in a response output last time. An information storage medium comprising information for recognizing a registered word included in a word input by a user from a registered word group to be selected. 12. An affirmative registration word for storing a judgment value for determining an evaluation value of a conversation situation in a current simulated conversation according to any one of claims 9 to 11, wherein said registration word means an affirmative response. And a negative registered word meaning a negative response, and a polysematic registered word meaning a polysematic response that can be taken either positively or negatively, and stored. When the registered word included in the word input by the above is recognized as the ambiguous registered word, the ambiguous registered word is changed to the positive registered word or the negative registered word according to the evaluation value determined by the determination information. An information storage medium characterized by including information for determining a response corresponding to a positive registered word or a negative registered word considered. 13. The information according to any one of claims 9 to 12, wherein: user profile information; and information for determining a response to be output based on the user profile information are stored. Storage medium. 14. The response according to claim 13, wherein the response is a response set as a word uttered by a virtual character whose profile is virtually set, and the decision information is a profile information of a user and a profile of the virtual character. An information storage medium including information for determining a response to be output based on a difference between the information storage medium and the information storage medium. 15. The method according to claim 14, wherein the plurality of replies are classified and stored in a group of replies corresponding to the assumed profile information of the user, and the decision information is a group of replies corresponding to the profile information of the user. An information storage medium including information for determining a response corresponding to a registered word from the inside. 16. The simulated conversation according to any one of claims 9 to 15, wherein the simulated conversation is a simulated conversation by voice, and stores voice data to be output as a reply, and the recognition information includes voice data input by a user. The information includes information for recognizing a registered word included in the plurality of registered words from among the plurality of registered words, and the output unit includes information for reading out audio data corresponding to the reply and outputting it as audio. Information storage medium.