JP2001188788A - Conversation processing apparatus and method, and recording medium - Google Patents

Abstract

(57) [Summary] [Problem] To provide a user with a variety of corresponding dialogues. SOLUTION: Information input to a language information input unit 41 by a user's voice or the like is subjected to language processing by a language processing unit 42. The dialogue management unit 43 creates data as a source of a dialogue sentence based on the processing result of the language processing unit 42 and the parameters stored in the parameter storage unit 44, and outputs the data to the response generation unit 45. The response generation unit 45 generates a response sentence based on the data from the interaction management unit 43 and according to the instruction of the reaction generation unit 47, and utters the response sentence. The parameters stored in the parameter storage unit 44 are dynamically updated by the parameter update unit 46 under various conditions.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a conversation processing apparatus and method, and a recording medium, and more particularly to a conversation processing apparatus and a conversation processing method for dynamically changing a parameter value to dynamically change a way of responding to a conversation. The present invention relates to a method and a recording medium.

[0002]

2. Description of the Related Art Hitherto, as toys and the like, a large number of robots (including stuffed ones) that output a synthetic sound when a touch switch is pressed have been commercialized.
In addition, a conversation with a computer is becoming popular.

[0003] In addition, a standard conversation system (task-oriented) with a computer is used, and there are services such as reservation of air tickets and provision of a travel guide service. This does not mean a natural conversation with a person, including a chat, only by talking about a predetermined matter. Eliza (James Allen: Natura) tried to make natural conversations, including chats, between computers and humans.
l Language Understanding, disclosed on pages 6-9).

[0004]

The above-mentioned Eliza and others describe that the contents of conversations with humans (users) are as follows.
He hardly understands, in other words, he can only talk back like a parrot, so that there is a problem that the user gets tired immediately.

[0005] The present invention has been made in view of such a situation, and by changing the parameter value dynamically, it is possible to bring about a change in the way of responding to conversation, and the like.
Accordingly, an object of the present invention is to provide a conversation that the user does not get tired of.

[0006]

According to a first aspect of the present invention, there is provided a conversation processing apparatus comprising: input means for inputting linguistic information; analysis means for analyzing linguistic information input by the input means; and storage means for storing parameters. And updating means for updating parameters stored in the storage means, and control means for controlling a conversation with the user based on the analysis result by the analysis means and the parameters stored in the storage means. Features.

[0007] The input means is provided by a user in a voice format,
At least one of a text format and a gesture format can be used to input language information supplied.

[0008] The updating means may start updating the parameters stored in the storage means in response to a user's voice instruction.

[0009] The updating means may include at least one of prosodic information obtained from the utterance of the user, information on the number of conversations with the user, and information on the time required until a response from the user is returned. Information, and that information is
When the predetermined condition is satisfied, the parameters stored in the storage unit can be updated.

[0010] The parameters may indicate the degree of talk, the degree of fidelity, the degree of interest, and the degree of correspondence, respectively.

According to a sixth aspect of the present invention, there is provided a conversation processing method, wherein an input control step of controlling input of linguistic information, an analyzing step of analyzing the linguistic information input in the processing of the input control step, and controlling storage of parameters. A conversation with the user is performed based on the storage control step, an update step of updating the parameters stored in the processing of the storage control step, an analysis result by the processing of the analysis step, and the parameters stored in the processing of the storage control step. And a control step of controlling.

According to a seventh aspect of the present invention, a program for a recording medium controls an input control step of controlling input of linguistic information, an analyzing step of analyzing linguistic information input in the processing of the input control step, and controls storage of parameters. A storage control step, an update step of updating parameters stored in the processing of the storage control step, a conversation with the user based on an analysis result by the processing of the analysis step, and the parameters stored in the processing of the storage control step. And a control step of controlling

[0013] The conversation processing apparatus according to claim 1, claim 6,
In the conversation processing method described in the above, the input language information is analyzed, the parameters are stored, the stored parameters are dynamically updated, the analysis result, The conversation with the user is controlled based on the stored parameters.

[0014]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing the appearance of a dialogue apparatus to which the present invention is applied. A keyboard 3 and a mouse 4 used when performing character input are connected to the main body 2 of the interactive device 1. Further, a microphone 5 used when performing voice input and a speaker 6 used when performing voice output are also connected. Further, a display 8 for displaying text and the like and a CCD (Charge Coupled Device) 8 used for capturing an image are also connected.

FIG. 2 is a diagram showing the internal configuration of the interactive device 1. The CPU 11 of the interactive device 1 has a ROM (Read Only Memor
y) Various processes are executed in accordance with the program stored in 12. A RAM (Random Access Memory) 13 stores data and programs necessary for the CPU 11 to execute various processes as appropriate. The input / output interface 15 includes an input unit 16 including a keyboard 3, a mouse 4, a microphone 5, and a CCD 8,
Output unit 17 composed of speaker 2 and display 7
The storage unit 18 composed of a hard disk (HDD) and the like, and a communication unit 19 for communicating with other devices via the Internet or the like are also connected to the input / output interface 15.

Further, the input / output interface 15 includes:
Magnetic disk 31, optical disk 32, magneto-optical disk 3
3, and a drive 20 for recording data on the semiconductor memory 34 or reproducing data.

FIG. 3 is a functional block diagram of the interactive device 1. The linguistic information input unit 41 includes a voice format, a text format,
It has a function of recognizing data input in an image format and converting it into linguistic information (text data). The language processing unit 42 analyzes the syntax and meaning of the language information output from the language information input unit 41, and detects a command or the like for controlling the interactive device 1.

The dialog management section 43 stores prosody information as a result of speech recognition output from the linguistic information input section 41, information on analysis results and information output from the language processing section 42, information on commands, and a parameter storage section 44. Based on the stored values of the various parameters, a response corresponding to the user's input is determined, and the response generator 45 outputs information for executing the determined response.

The parameter updating unit 46 analyzes the language information obtained by the user,
It has a function of updating parameter values stored in the parameter storage unit 44 using prosody information and information on various parameters. The reaction generation unit 47 determines the value of various parameters stored in the parameter storage unit 44,
Determine the response to the user. The reaction generated by the reaction generation unit 47 includes, for example, changing the sound output from the speaker 6 and changing the expression of the character displayed on the display 7.

The response generation unit 45 responds to the response command output from the dialog management unit 43, the content of the reaction determined by the reaction generation unit 47, and the values of various parameters stored in the parameter storage unit 44. And a response of the character, and output a response sentence as a voice through the speaker 6 or change the display of the character displayed on the display 7.

FIG. 4 is a diagram showing the internal configuration of the language information input unit 41. The user can interact with the interactive device 1 using voice, text, and gesture. The fact that the user has spoken is taken into the language information input unit 41 by the microphone 5. Microphone 5
Is input to the voice recognition unit 51, and voice recognition is performed. The image of the user captured by the CCD 8 is regarded as a gesture, and is recognized by the gesture recognition unit 52. The recognition result (text data) output from the voice recognition unit 51, the keyboard 3, or the gesture recognition unit 52 is output to the language processing unit 42.

FIG. 5 is a diagram showing the internal configuration of the voice recognition unit 51. The utterance of the user is converted into a voice signal via the microphone 5 by an AD (Analog Digita) of the voice recognition unit 51.
l) Input to the conversion unit 61. The AD converter 61 samples and quantizes the audio signal, which is an analog signal,
It is converted into audio data which is a digital signal. This audio data is supplied to the feature extraction unit 62.

The feature extraction unit 62 converts the audio data from the AD conversion unit 61 into appropriate frames, for example,
A feature parameter such as a spectrum, a linear prediction coefficient, a cepstrum coefficient, and a line spectrum pair is extracted and supplied to the matching unit 63. In addition, the prosody information is output from the feature extraction unit 62 to the dialog management unit 43 and the like as necessary.

The matching unit 63 is based on the feature parameters from the feature extraction unit 62 and
4. Dictionary database 65 and grammar database 6
The input voice signal is recognized while referring to 6 as necessary.

That is, the acoustic model database 64 stores acoustic models representing acoustic features such as individual phonemes and syllables in the language of the speech to be recognized. Here, as the acoustic model, for example, HMM (Hidden Markov
Model) can be used. Dictionary database 6
Reference numeral 5 stores a word dictionary in which information on pronunciation of each word to be recognized is described. The grammar database 66 stores grammar rules that describe how the words registered in the word dictionary of the dictionary database 65 are linked (connected). Here, as the grammar rule, for example, a rule based on a context-free grammar (CFG), a statistical word chain probability (N-gram), or the like can be used.

The matching section 63 includes a dictionary database 6
By connecting the acoustic models stored in the acoustic model database 64 by referring to the 5th word dictionary, an acoustic model (word model) of a word is formed. Further, the matching unit 63 connects several word models by referring to the grammar rules stored in the grammar database 66, and uses the word models connected in this way, based on the feature parameters, for example, The input speech is recognized by the HMM method or the like. Then, the speech recognition result by the matching unit 63 is output as, for example, text data.

Although the detailed configuration of the gesture recognition unit 52 is not shown, it is determined whether or not the user has performed a predetermined operation by using a method such as template matching, and if it is determined that the user has performed the predetermined operation, The text data corresponding to the operation is output as a recognition result.

The text data input by the user using the keyboard 3 is used as a recognition result. Keyboard 3, voice recognition unit 51, or gesture recognition unit 5
2 is output to the language processing unit 42. FIG. 6 is a diagram showing the internal configuration of the language processing unit 42.

The recognition result (text data) output from the language information input unit 41 is input to the language analysis unit 71 of the language processing unit 42. The language analysis unit 71, based on data stored in the dictionary database 72, the grammar database for analysis 73, and the command dictionary database 74,
The input speech recognition result is analyzed such as morphological analysis and syntax analysis to extract language information and command information such as word information and syntax information. Further, based on the contents described in the dictionary, the meaning and intention of the input utterance are also extracted.

That is, the dictionary database 72 stores information such as part-of-speech information necessary for applying word notation and analysis grammar, and individual semantic information of words, and is stored in the analysis grammar database 73. Stores data describing restrictions on word chains based on information on each word stored in the dictionary database 72. The command dictionary database 74 stores data on commands. Using these data, the language processing unit 42 (language analysis unit 71) analyzes the text data of the input speech recognition result.

The data stored in the analysis grammar database 73 includes regular grammar, context-free grammar, statistical word chain establishment, and language theory including semantics such as HPSG when including semantic analysis. Is data necessary for text analysis.

The analysis result output from the language processing unit 42 is supplied to the dialog management unit 43. FIG. 7 is a diagram showing the internal configuration of the dialog management unit 43. The dialogue processing unit 81 of the dialogue management unit 43 uses the input analysis result to create data on which a response sentence generated by the response generation unit 45 at the subsequent stage is based. The dialogue processing unit 81 includes a scenario (details will be described later) stored in the scenario memory 82, knowledge stored in the knowledge database 83, and the dialogue history memory 8
Reference is made to the history stored in No. 4 to create data serving as the basis of the response sentence.

The scenario stored in the scenario memory 82 describes a dialogue pattern, and differs depending on the task. For example, a scenario related to a goal-oriented task (parameter setting) is a scenario in which a user is asked a question and information is collected. An example is shown below. (Action (Question (speak, devotion, interest, reciv
e))) (speak ???) # chatter (devotion ???) # fidelity (interest ???) # interest (recive ???) # reception

As another example, a non-purpose type (eg, ELAZA) dialogue scenario (used for chat, etc.) is as follows. If X exist then speak (Y) #X: keyword, Y: response sentence (X: money Y: what do you want) # (XY) (X: I want to eat Y: I am hungry)

As the knowledge stored in the knowledge database 83, control information relating to the interactive operation and information necessary for a response are managed. For example, there are rules in a dialog such as returning a greeting response when it is determined that the user has made a greeting, a topic used in a chat mode (task), and the like.

The dialog history memory 84 stores various information (for example, user input, response content, etc.) exchanged between the user and the dialog device 1 and dialog management information (for example, the number of responses, response time, etc.). It is remembered. These pieces of information stored in the dialog history memory 84 are used for dialog processing such as generation of a response sentence and control of the number of responses.

The parameter storage section 44 stores parameters as shown in FIG. FIG. 8A shows a case where the values of the parameters are represented as discrete values, and FIG.
Indicates a case where the value of the parameter is represented as a continuous value. In an actual configuration, either one of a discrete value or a continuous value may be used. The values of these parameters are updated by an instruction from the parameter updating unit 46.

The parameters stored in the parameter storage unit 44 control the selection and generation of a response sentence and the change in voice pitch when responding to a user input, for example, when responding by voice. , And is also used to control changes in the expression of the character displayed on the display 7.

The identifier is a symbol (information) for distinguishing each parameter, the type represents the characteristic of the parameter, and the value represents the degree of influence of each parameter. The probability value is used at the time of updating the parameter,
When the value is represented by a discrete value as in (A), the value is 0.
Alternatively, when the values are represented by continuous values as shown in FIG. 8B, the probability is that each value increases or decreases.

The parameters stored in the parameter storage unit 44 are updated according to an instruction from the parameter update unit 46. The parameter updating unit 46 performs, for example, according to a processing model as shown in FIG. Each neuron (N1 to N4) is connected by an interconnected network, and an input to a given neuron changes its own value or the value of another neuron.
As the type of the connection, a side excitatory connection, a side suppression connection, or the like can be used.

The reaction generator 47 holds a table as shown in FIG. That is, among the parameter information stored in the parameter storage unit 44 shown in FIG. 9, the identifier, the type, and the condition (in the example shown in FIG. 10, corresponding to the value in FIG. 8B), This is a table for determining the content of the reaction.

FIG. 11 is a diagram showing the internal configuration of the response generation unit 45. The response sentence creation data output from the dialog management unit 43 is input to the response sentence generation unit 91 of the response generation unit 45. The response sentence generation unit 91 generates an appropriate response sentence (text data) by referring to data stored in the template database 92, the response sentence generation rule database 93, and the dictionary database 94.

The template database 92 stores template information. For example, when a continuous value is used as a parameter value, the template information is as follows. Reaction condition Sentence example R3 7 or more Careful (-I will do it) R3 4 or more (-I do) A template is information that determines what kind of attitude a response sentence is generated based on an identifier and a condition.

The response sentence generation rule database 93 stores rules regarding what response sentence is generated and how. For example, if topics are managed in a semantic frame structure, rules for generating a natural language sentence therefrom are written. The method of generating a natural language sentence from this semantic structure can be realized by performing the reverse of the processing performed by the language processing unit 42. The dictionary database 94 stores word information necessary for creating a response sentence.

Thus, the response sentence as text data generated by the response sentence generation unit 91 is output to the text analysis unit 96. The text analysis unit 96 analyzes the input text data with reference to the dictionary database 94 and the grammar database 95 for analysis.

That is, the dictionary database 94 stores a word dictionary in which part-of-speech information of each word, information such as reading, accent, and the like are described. For words described in the word dictionary, grammatical rules for analysis such as restrictions on word chains are stored. Then, the text analysis unit 9
6 is based on the word dictionary and the grammatical rules for analysis,
An analysis such as morphological analysis or syntax analysis of the text input thereto is performed, and information necessary for rule-based speech synthesis performed by the rule synthesis unit 97 at the subsequent stage is extracted. Here, the information necessary for the rule speech synthesis includes, for example, information for controlling the position of a pause, accent and intonation, other prosody information, and phonological information such as pronunciation of each word.

The information obtained by the text analysis unit 96 is supplied to the rule synthesis unit 97, and the rule synthesis unit 97 uses the phoneme segment database 98 to generate synthesized speech corresponding to the text input to the text analysis unit 96. (Digital data) is generated.

That is, the speech segment database 98 stores speech segment data in the form of, for example, CV (Consonant, Vowel), VCV, CVC, and the like. Based on the information, necessary phoneme segment data is connected, and a pause, an accent, an intonation, and the like are appropriately added, so that the text analysis unit 96 can be used.
To generate speech data of a synthesized sound corresponding to the text input to the.

The audio data is supplied to a DA converter 99, where it is converted into an audio signal as an analog signal. This audio signal is supplied to the speaker 6 (FIG. 1), whereby a synthesized sound corresponding to the text input to the text analysis unit 96 is output.

The operation of the interactive device 1 will be described below. First, in response to a user's instruction, a program for performing a dialogue stored in the storage unit 18 (FIG. 2) is started, and a screen displayed when the dialogue is performed is displayed on the display 7 (FIG. 1). . FIG. 12 shows a display example of a screen displayed on the display 7. As shown in FIG. 12, on the display 7, a character display section 11 is provided.
1, text display unit 112, and parameter setting unit 1
13 is displayed.

The character display section 111 displays a character selected by the user from among a plurality of characters registered in advance, and as described above, the expression and attitude change according to the value of the parameter. The text display unit 112 displays, as text, the fact that the user has spoken or that the interactive device 1 has spoken. As the text displayed on the text display unit 112, the fact that the user has made an input with the keyboard 3 may be displayed, and further, a response sentence of the interactive device 1 to the user input may be displayed. Either a sentence input by the user or a response sentence of the interactive device 1 may be displayed.

It is also possible to display the recognition result for the user's utterance input from the microphone 5 and display the response sentence of the interactive device 1 to the recognition result. Also in this case, either one of the sentence input by the user by voice or the response sentence of the interactive device 1 may be displayed. Further, the response sentence of the interactive device 1 may be displayed on the text display unit 112 and the response sentence having the same content may be uttered from the speaker 6. Further, when making the speaker 6 speak, the text display unit 112
, The response sentence of the interactive device 1 may not be displayed.

Further, the recognition result regarding the user's gesture input by the CCD 8 may be displayed.

If what can be displayed and what should not be displayed can be set by the user, it is possible to provide the user-friendly interactive device 1.

The parameter setting section 113 is a section for changing the parameters stored in the parameter storage section 44, and thus executing the function of the parameter updating section 46. FIG. 13 is a diagram illustrating details of the parameter setting unit 113. In the example shown in FIG. 13, ""
Four parameters can be set: “talking degree”, “fidelity”, “interest level”, and “agreement level.” A check box is provided on the left side of each parameter. The value of the checked parameter is changed.

Here, the parameter updating process performed by the parameter updating unit 46 will be described with reference to the flowchart of FIG. The parameter updating unit 46 determines in step S
At 1, it is determined whether or not the update instruction is issued by the user. The case where the user instructs to update the parameter may be when the check box of the parameter setting unit 113 is checked (removed) or when an instruction is given by a voice command. Therefore, if it is determined in step S1 that the instruction is an update instruction by the user, the process proceeds to step S2, and it is determined whether the instruction is an update instruction by a voice command.

If it is determined in step S2 that the instruction is not an update instruction by a voice command, in other words, if it is determined that the check box (FIG. 13) of the parameter setting unit 113 has been operated, the process proceeds to step S3 and the setting is performed. The update process by the unit is started. FIG. 15 is a flowchart illustrating a detailed process of step S3. Step S2
In step 1, by determining which of the check boxes of the parameter setting unit 113 is turned on or off, the parameter for which updating is instructed is determined.

When a discrete value (0 or 1) is used as a parameter, when a check box is turned on, the turned-on parameter is set to 1, and when turned off, the turned-off parameter is turned off. May be set to 0. However, in such a process that faithfully follows the user's instruction, for example, the interactive device 1 always responds the same to a user who does not change the setting of the parameter setting unit 113 at all. . When a continuous value is used as a parameter value, the value does not change. Therefore, in the present embodiment, there is provided a step of determining whether or not to change the parameter value even if the instruction is from the user.

In step S22, a parameter updating process is performed. FIG. 16 is a flowchart illustrating the detailed process of step S22. In step S31, a change processing calculation using a probability value is performed. For example, if the “talking degree” check box of the parameter setting unit 113 shown in FIG. 13 is turned off, first, the parameter updating unit 46 sets the parameter information related to “talking degree” stored in the parameter storage unit 44. Is read. In this case, assuming that a discrete value expression is used, the value of the parameter of "talking degree" is 1 and the probability value is 0.5 according to FIG. 8A.

In this case, since the value of the parameter of "talking degree" is 1, it is updated to 0 with a probability of 1/2. In step S31, it is determined whether or not to change the value of the parameter in step S32 by using the result of performing the change processing calculation using the probability value.
When it is determined in step S32 that the value of the parameter is to be changed, the parameter is changed in step S34 according to the determination. On the other hand, step S
If it is determined in step 32 that the parameter value is not to be changed, the parameter change process is not performed and step S
Returning to the process of FIG. 1 (FIG. 14), the subsequent processes are performed.

Returning to the description of the flowchart of FIG. 14, when it is determined in step S2 that the update instruction is a voice command, the process proceeds to step S4, and the voice update process is started. When it is determined that the instruction is an update instruction by a voice command, the user utters “update parameter”, the speech recognition processing is executed by the language processing input unit 41, and the syntax analysis and the like are performed by the language processing unit 42. In this case, the command is analyzed as a command. The dialogue processing unit 43 receives an analysis result indicating that the command is a command from the language processing unit 42 and, when determining that the command is a parameter update instruction, issues a parameter update instruction to the parameter update unit 46. As described above, by receiving the update instruction, the parameter update unit 46
It is determined in step S2 whether or not the update instruction is a voice command.

FIG. 17 is a flowchart illustrating the details of the process in step S4. In step S41, a parameter update process is performed. Since the details of the parameter update processing have already been described with reference to the flowchart of FIG. 16, the description thereof is omitted here. While the parameter updating process is being performed by the parameter updating unit 46, in step S42, the dialog management unit 43 reads a scenario corresponding to the input command from the scenario memory 82 (FIG. 7).

The dialogue processing unit 43 asks the user whether or not there is any other parameter update based on the read scenario and whether or not the parameter update processing can be ended.
Upon receiving the reply, the processing corresponding to the reply is performed based on the scenario. By performing such processing, it is determined whether or not the user has instructed the end of the parameter update processing. If it is determined that the user has not instructed, the process proceeds to step S41.
And the same processing is repeated for the other parameters. On the other hand, if it is determined in step S43 that the processing is to be ended, the process returns to step S1 (FIG. 14), and the subsequent processing is repeated.

Here, the above-described processing will be described with a specific example. At time t _1, the user, to the utterance, "I want to change the personality of the character." Upon receiving the utterance, the interactive device 1 sets the language input unit 41, the language processing unit 42,
By performing the above-described processing in each unit of the dialogue processing unit 43, the dialogue management unit 43 determines that the user wants to change the character of the character and that it is instructed to update the parameter value.

Then, the dialogue management unit 43 reads a scenario related to parameter update from the scenario memory 82, and according to the scenario, at time t ₂ , writes a message such as “Update parameters. Which parameter should be changed?” Generate data for speaking. Then, using the generated data, the response generation unit 4
5 and the speaker 6. Response to such utterances, the user At time t _3, utters "chatter degree increased to". In response to the utterance, the parameter updating unit 46 updates the value of the parameter related to “talking degree” stored in the parameter storage unit 44 according to the above-described processing.

Further, at time t ₄ , the interactive device 1 asks a question “I have increased the degree of talk. What else should I do?”. In contrast, at time t _5, when the user utters "fidelity raised", updating of the parameters of "fidelity" is carried out by the same process as in the "chatter" specification, users'"When speaking with, interactive processing apparatus 1, it is determined that the instruction of termination is issued, at time t _6," the other as it is to speak and by the end of the process of updating the parameter ".

As described above, the parameter updating process can be performed by interacting with the user.

Returning to the description of the processing of the flowchart in FIG. 14, when it is determined in step S1 that the instruction is not an update instruction by the user, the process proceeds to step S5, and it is determined whether the instruction is an update instruction based on prosody information. If it is determined that the update instruction is based on the prosody information, the voice recognition unit 51
The prosody information extracted by the process of the feature extracting unit 62 (FIG. 5) is input to the dialog managing unit 43, and the dialog managing unit 43
Refers to the knowledge stored in the knowledge database 83 (FIG. 7) and obtains processing information related to prosodic information. Then, based on the acquired processing information, it is determined whether the instruction is an update instruction.

If it is determined that the update instruction is based on the prosody information, the process proceeds to step S6, and the update process based on the prosody information is started. FIG. 18 is a flowchart illustrating details of the process in step S6. Here, a case where a pitch is used as the prosody information will be described as an example. In step S51, the dialog management unit 43 determines whether or not the pitch is high based on the input prosody information. If it is determined in step S51 that the pitch is high, the process proceeds to step S53, where the parameter updating unit 4
At 6, an instruction is issued to update in the direction of increasing the value of the "fidelity" parameter.

On the other hand, if it is determined in step S51 that the pitch is not high, the process proceeds to step S53,
The parameter updating unit 46 is instructed to update in the direction of decreasing the value of the parameter of “fidelity”.

When such processing is completed, the process returns to step S1, and the subsequent processing is repeated.

If it is determined in step S5 (FIG. 14) that the instruction is not an update instruction based on prosody information, the flow advances to step S7 to determine whether the instruction is an update instruction based on internal information. The internal information is, for example, a time until the user responds to the speech of the interactive device 1 or the number of interactions between the interactive device 1 and the user. These pieces of information are stored in the dialog history memory 84 (FIG. 7). The dialogue management unit 43 refers to the information stored in the dialogue history memory 84 and satisfies predetermined conditions such as the above-described time until a reply exceeds a predetermined value and the number of conversations exceeds a predetermined value. It is determined that an update instruction based on the internal information has been issued.

If it is determined in step S7 that an update instruction based on the internal information has been issued, the process proceeds to step S8, and an update process based on the internal information is started. FIG.
It is a flowchart explaining the detail of a process of step S8. Here, a case will be described in which the time until the user returns a response (user response time) to the utterance of the interactive device 1 is used as the internal information. In step S61, the dialog management unit 43 determines whether the user response time is short. The determination as to whether or not it is short is made using a preset value as a threshold, and if it is equal to or less than the threshold, it is determined to be short.

If it is determined in step S61 that the user response time is short, the process proceeds to step S62, and the parameter updating unit 46 is instructed to update the parameter of the "degree of interest" in a direction of increasing the value. . on the other hand,
If it is determined in step S61 that the user response time is not short, the process proceeds to step S63, and the parameter updating unit 46 is instructed to update the parameter value of “degree of interest” in a decreasing direction.

When such processing is completed, the flow returns to step S1, and the subsequent processing is repeated.

If it is determined in step S7 that the instruction is not an update instruction based on internal information, the process proceeds to step S9. In step S9, it is determined whether the instruction is a neural update instruction. This determination is made, for example, by setting a predetermined cycle and at each cycle. If it is determined that the instruction is an update instruction by the neural network, the process proceeds to step S10, and the update process by the neural network is started. FIG. 20 is a flowchart illustrating details of the process in step S10. As the neural network, for example, the processing model shown in FIG. 9 is used.

In step S 71, the parameter updating section 46 acquires the value of the parameter stored in the parameter storage section 44. Then, in step S72, the acquired parameter values are input to the neuron corresponding to each parameter. The calculation of each parameter is performed using the neural network, and the value of each parameter updated from each neuron is calculated.

In step S73, the parameter values stored in the parameter storage unit 44 are updated using the parameter values calculated in step S72. When such processing is completed, step S
The process returns to 1 and the subsequent processes are repeated.

The processing of the flowchart shown in FIG. 14 is repeatedly performed, and the parameter values stored in the parameter storage unit 44 are dynamically updated. in this way,
By updating the parameters by the user's instruction and the processing of the interactive device 1 itself, the interactive device 1 for the user is updated.
Can be changed, and a natural and varied conversation can be provided to the user.

FIG. 21 shows a process of generating a response sentence of the interactive apparatus 1 using such dynamically changing parameters.
This will be described with reference to the flowchart of FIG. Reaction generator 47
Reads the value of each parameter stored in the parameter storage unit 44 in step S81. Based on the read value, the process from step S82 is performed. First, in step S82, a process of "talking degree" is performed. FIG. 22 is a flowchart illustrating details of the process in step S82.

In step S91, a reaction is selected based on the read parameter values. Reaction generator 4
7 holds a table in which the relationship between parameter values and reactions as shown in FIG. 10 is described, as described above. The reaction is selected according to the table, but the type of the chatter is P1 (corresponding to the identifier in the table shown in FIG. 8), and the response is "chat" or "the number of words is small.""Naru" is set. Either of the two reactions is selected depending on the conditions.

That is, when a continuous value is used as a parameter value, if the value is 5 or more, "chat" is selected, and if it is 5 or less, "the number of words is reduced" is selected. . In step S92, it is determined whether or not the reaction selected in step S91 is chat. If it is determined in step S92 that chat is selected, the process proceeds to step S93. In step S93, the interaction management unit 43 determines the number of responses with the user according to the value of the parameter. The larger the value of the parameter, that is, the higher the chatter degree, the larger the value of the number of responses is set.

In step S94, the dialog management section 81
Reads the chat mode scenario stored in the scenario memory 82, and in step S95, speaks to the user under the initiative of the system (interactive device 1). As described above, when the chatter degree is high, the interactive device 1 talks to the user.

On the other hand, in step S92, it is determined that chat is not selected, ie,
If it is determined that the chatter degree is not high and the number of words is small but selected, the dialogue device 1 is less likely to talk to the user.
Step 95 is skipped.

When the process relating to the chatter degree is completed in this way, the process proceeds to step S83 (FIG. 21),
Fidelity processing begins. FIG.
5 is a flowchart for explaining the details of the processing of FIG. In step S101, the reaction generation unit 47 reads a parameter value related to fidelity stored in the parameter storage unit 44. The reaction generation unit 47 holds the
Refer to the table as shown in. As the response regarding the fidelity whose type is P2, R3 to R3 are used as identifiers.
Five reactions are set.

In step S102, the reaction generator 47 determines which condition the parameter value relating to fidelity read from the parameter storage unit 44 applies to, and selects a reaction of the applicable condition. For example, when the value of the parameter is 8, the condition of 7 or more is satisfied, and therefore, the reaction of the identifier R3 of "process as instructed by the dialogue" is selected.

Based on the reaction thus selected,
In step S103, the response generation unit 45 is controlled so that an appropriate response sentence is created. For example, when "process as directed by dialog" of the identifier R3 is selected as the response, the response sentence generating unit 91 of the response generating unit 45 refers to the template stored in the template database 92, and in the case of the identifier R3. , And generates a response sentence using a polite expression (-Itashimashika).

When the processing of such fidelity is completed,
Proceeding to step S84 (FIG. 21), interest level processing is performed. FIG. 24 is a flowchart illustrating details of the process in step S84. In step S111, the reaction generation unit 47 reads the parameter value related to the degree of interest stored in the parameter storage unit 44. The reaction generation unit 47 refers to the held table and recognizes that the type of the reaction is P3 and that two identifiers R6 and R7 are set as the reactions corresponding to the type.

In step S112, it is determined whether or not the number of companions is large but selected. In other words,
It is determined whether the value of the read parameter is 6 or more. It is determined in this way that if the value of the parameter is 6 or more, “the number of companions” of the identifier R6 is selected as the response, and if the value of the parameter is 6 or less, the response is This is because "listening" of the identifier R7 is selected.

If it is determined in step S112 that the number of companions is large but selected, the process proceeds to step S11.
Proceeding to 3, the dialogue management unit 43 sets the number of hits for the user's utterance and the number of hits. On the other hand, if it is determined in step S112 that the number of companions is large but not selected, in other words, if it is determined that “listen” is selected as the response, the number of companions is set. Since there is no need to perform the process, the process of step S113 is skipped.

When the processing of the degree of interest is completed,
Proceeding to step S85 (FIG. 21), the processing of the response level is started. FIG. 25 is a flowchart illustrating details of the process in step S85. In step S121, the reaction generation unit 47 reads the value of the parameter related to the degree of response stored in the parameter storage unit 44. In step S122, prosody information is calculated.

The reaction generator 47 refers to the table held and recognizes that the type of the reaction is P4, and two identifiers R8 and R9 are set as the reactions corresponding to the type. That is, as the reaction regarding the degree of correspondence, two settings are made for the identifier R8, "Expression and voice are bright," and for the identifier R9, "Expression is dark and angry." This reaction is performed by the character display unit 111 (FIG. 1).
This relates to the expression of the character displayed in 2) and the voice of the interactive device 1 uttered by the speaker 6.

Here, the processing relating to the voice will be described as an example. In step S122, prosody information which determines the brightness and darkness of the voice is calculated. Based on the calculation result, in step S123, a voice is synthesized by a voice synthesis unit (not shown) and output.

The processing of the flowchart shown in FIG.
It is repeatedly performed while the interactive device 1 is interacting with the user. Since the value of the parameter is dynamically changed as described above, by repeatedly performing the processing of the flowchart shown in FIG. 21 during the dialogue, it is possible to bring about a change in the way the dialogue apparatus 1 responds. Therefore, it is not necessary to perform a monotonous dialog, and it is possible to provide a fun dialog to the user.

The above series of processing can be executed by hardware, but can also be executed by software. When a series of processing is executed by software, it is possible to execute various functions by installing a computer in which the programs constituting the software are embedded in dedicated hardware, or by installing various programs For example, it is installed from a recording medium to a general-purpose personal computer or the like.

As shown in FIG. 2, the recording medium is a magnetic disk 31 (including a floppy disk) on which the program is recorded and an optical disk 32 which are distributed separately from the computer to provide the program to the user.
(CD-ROM (Compact Disk-Read Only Memory), DVD (Dig
ital Versatile Disk), magneto-optical disk 33
(Including MD (Mini-Disk)) or semiconductor memory 3
ROM 12 that stores programs and is provided to the user in a state in which it is not only configured by a package medium such as
And a hard disk including the storage unit 18.

In the present specification, the steps of describing a program provided by a medium are not limited to processing performed in chronological order according to the order described, but are not necessarily performed in chronological order. Alternatively, it also includes individually executed processing.

[0098] In this specification, the system is
It represents the entire device composed of a plurality of devices.

[0099]

As described above, according to the conversation processing apparatus of the first aspect, the conversation processing method of the sixth aspect, and the recording medium of the seventh aspect, the input language information is analyzed. The parameters are stored, the stored parameters are dynamically updated, and the conversation with the user is controlled based on the analysis result and the stored parameters. It is possible to provide a corresponding dialogue.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an appearance of an embodiment of a dialog device 1 to which the present invention is applied.

FIG. 2 is a diagram showing an internal configuration of the interactive device 1.

FIG. 3 is a functional block diagram of the interactive device 1.

FIG. 4 is a diagram showing an internal configuration of a language information input unit 41.

FIG. 5 is a diagram showing an internal configuration of a voice recognition unit 51.

FIG. 6 is a diagram showing an internal configuration of a language processing unit 42.

FIG. 7 is a diagram showing an internal configuration of a dialog management unit 43.

FIG. 8 is a diagram illustrating a table stored in a parameter storage unit 4;

FIG. 9 is a diagram illustrating parameter updating performed by a parameter updating unit.

FIG. 10 is a diagram illustrating a table stored in a reaction generation unit 47.

11 is a diagram showing an internal configuration of a response generation unit 45. FIG.

FIG. 12 is a display example displayed on a display 7;

13 is a display example of a parameter setting unit 113. FIG.

FIG. 14 is a flowchart illustrating a parameter update process.

FIG. 15 is a flowchart illustrating details of a process in step S3 of FIG. 14;

FIG. 16 is a flowchart illustrating details of a process in step S22 in FIG. 15;

FIG. 17 is a flowchart illustrating details of a process in step S4 of FIG. 14;

FIG. 18 is a flowchart illustrating details of a process in step S6 of FIG.

FIG. 19 is a flowchart illustrating details of a process in step S8 in FIG. 14;

FIG. 20 is a flowchart illustrating details of a process in step S10 of FIG. 14;

FIG. 21 is a flowchart illustrating a response sentence generation process.

FIG. 22 is a flowchart illustrating details of a process in step S82 in FIG. 21;

FIG. 23 is a flowchart illustrating details of a process in step S83 of FIG. 21;

FIG. 24 is a flowchart illustrating details of a process in step S84 in FIG. 21;

FIG. 25 is a flowchart illustrating details of a process in step S85 in FIG. 21;

[Explanation of symbols]

1 interactive device, 2 main body, 3 keyboard, 4
Mouse, 5 microphones, 6 speakers, 7
Display, 8 CCD, 41 language information input unit, 42 language processing unit, 43 dialogue management unit, 44
Parameter storage unit, 45 response generation unit, 46 parameter update unit, 47 reaction generation unit

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 15/22 G10L 3/00 571U (72) Inventor Sachi Tanaka 6-7-35 Kita Shinagawa, Shinagawa-ku, Tokyo No. Sony Corporation (72) Inventor Jun Yokono 6-35, Kita Shinagawa, Shinagawa-ku, Tokyo Sony Corporation (72) Inventor Toshio Oe 6-35, Kita-Shinagawa, Shinagawa-ku, Tokyo Soniー Internal F term (Reference) 5B049 BB31 BB52 CC06 DD00 DD01 DD03 DD05 EE03 EE05 EE07 FF03 FF04 FF06 FF09 GG04 GG07 5B091 AA15 BA03 BA19 CA02 CA05 CA12 CB12 CB22 CB32 CC02 CC03 CD03 CD15 EA01 EA21 KK02 AB11 9A001 BB02 BB03 BB04 CC02 DD11 EE02 EE05 HH06 HH07 HH11 HH16 HH17 HH18 HH21 HH33 JJ01 JJ62 JJ72 KK46 KK56 KK60

Claims (7)

[Claims] 1. A conversation processing apparatus for talking with a user, wherein: input means for inputting linguistic information; analysis means for analyzing the linguistic information input by the input means; storage means for storing parameters; Updating means for updating the parameters stored in the means; control means for controlling a conversation with the user based on an analysis result by the analyzing means and the parameters stored in the storing means. A conversation processing device characterized by the above-mentioned. 2. The method according to claim 1, wherein the input unit inputs the language information supplied from the user using at least one of a voice format, a text format, and a gesture format. The conversation processing device according to claim 1. 3. The conversation processing apparatus according to claim 1, wherein the updating unit starts updating the parameters stored in the storage unit in response to a voice instruction from the user. 4. The method according to claim 1, wherein the updating means includes: prosody information obtained from the utterance of the user; information on the number of conversations with the user; and information on the time required for the user to return a response. 2. The conversation processing apparatus according to claim 1, wherein at least one piece of information is used, and when the information satisfies a predetermined condition, the parameter stored in the storage unit is updated. 5. The conversation processing apparatus according to claim 1, wherein the parameters indicate a degree of talk, a degree of fidelity, a degree of interest, and a degree of correspondence. 6. A conversation processing method of a conversation processing apparatus for conversation with a user, wherein: an input control step of controlling input of linguistic information; and an analyzing step of analyzing the linguistic information input in the processing of the input control step. A storage control step of controlling storage of the parameter; an update step of updating the parameter stored in the processing of the storage control step; an analysis result by the processing of the analysis step; and an analysis result stored in the processing of the storage control step. Based on the above parameters,
And a control step of controlling a conversation with the user. 7. A program for a conversation processing of a conversation processing apparatus for conversation with a user, comprising: an input control step of controlling input of linguistic information; and analyzing the linguistic information input in the processing of the input control step. An analysis step, a storage control step of controlling storage of parameters, an update step of updating the parameters stored in the processing of the storage control step, an analysis result by the processing of the analysis step, and a processing of the storage control step Based on the parameters stored in
And a control step of controlling a conversation with the user. A recording medium on which a computer-readable program is recorded.