JP3005249B2 - Fuzzy pattern type human interface system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial applications]

The present invention relates to computer games, computer picture books,
Characters with rich personality have appeared in software such as psychotherapy, CAI, and failure diagnosis systems, and through dialogue with the user, the characters can react to human situations and construct scene developments suited to the situation. A human interface system that can be used.

[Prior art]

Traditionally, like game computer software,
In the case of software that includes a large amount of image data and uses a large memory area for the image data, the memory area that can be used by the program is reduced. Because of the limited memory area that can be used, it has been difficult to create software that can display unique and complex characters and develop flexible stories. For example, in the case of a computer game, the personality of a character (character) has conventionally been differentiated by face, occupation, physical strength, etc., but the range of individuality is small, and it is not possible to express the psychological aspect of the character. Was. The dialogue was uninteresting because there were many uniform ones and lacked flexibility because only fixed story development was possible. In addition, there is no interest because the developing situation is displayed with a bland number such as a physical strength value. Since the game user can only participate in the game overlooking the whole game, he can only enjoy the game from a starting point from one direction.

[Problems to be solved by the invention]

The present invention solves the above-mentioned disadvantages of the prior art, and enables a complex character with a rich personality to appear even in a compact program that can use a limited memory area, thereby enabling a flexible story development. It is an object of the present invention to provide a human interface system that enables the above.

[Means for Solving the Problems]

The fuzzy pattern type human interface of the present invention includes an input unit (FIG. 101) capable of inputting an instruction or a question from an operator in the form of words, a target model having a plurality of models including a psychological model, and each model. A database storing fuzzy character personality data representing the character's personality by fuzzy pattern data corresponding to
A diagnostic unit (104, 105 in FIG. 1) for diagnosing the psychological situation of the character itself by collating the information of the instruction or question in words input from the input unit with the fuzzy pattern data of the fuzzy character personality data and inputting the character An action selection unit (FIG. 107) for selecting an action to be performed based on the obtained information and the diagnosis result of the diagnosis unit, and an action generation unit (FIG. 108) for generating an action corresponding to the selected action. It has. According to one aspect of the present invention, the diagnosis unit obtains a state of the character itself with respect to a word input to the character or a change in surrounding environment by comparing it with fuzzy pattern data, and outputs a diagnosis result. A situation diagnosis unit (FIG. 104). According to another aspect of the present invention, the diagnosis unit obtains a psychological effect of the input word on the character by comparing the psychological effect with the fuzzy pattern data based on the character's own state and the word. And a psychological mechanism processing unit (909 in FIG. 9) for generating a mood model representing a psychological state. Further, according to another aspect of the present invention, the diagnostic unit includes:
A mental information diagnosis processing unit that obtains a mental state with respect to the input instruction and the action of the character by comparing it with fuzzy pattern data based on the state of the character itself and outputs a diagnosis result (FIG. 9, 908) It has. Further, in another aspect of the present invention, a decision-making processing unit that receives the mood model and the diagnosis result, compares it with fuzzy character pattern data representing various mental states of the character, and decides on a matter to be decided. (9th
(Fig. 912).

[Action]

The target model has a plurality of models and represents the current state of the character. The database stores data representing the character's personality in a fuzzy pattern corresponding to each model. The diagnosis unit converts a target model representing the state of the character into a fuzzy pattern of character personality data limited by input (an example is shown in FIGS. 15 and 18).
And the state of the character is diagnosed. If the input is verbal information, the diagnosis unit diagnoses the psychological situation of the character itself by comparing it with fuzzy pattern data relating to psychology. The action selection unit selects an action to be performed based on the result of the diagnosis, and the action generation unit generates a corresponding action. According to the present invention, the personality of the character is set by the fuzzy pattern data of a plurality of models, so that the character characterization and action pattern of the character can be set to complex and various types. The model includes a psychological model, and can perform a self-diagnosis process on a verbal instruction, a question, and the like, so that a character having a human nature can be handled and an excellent human interface can be constructed. Since the present invention uses the fuzzy pattern data of a plurality of models as described above, the program capacity becomes extremely small, and even if the program is operated in a limited memory area, a flexible story can be developed with complicated characters. .

【Example】

[1] Schematic Configuration of Embodiment FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention.
It is. When specifically explaining and explaining in this embodiment, “Mota
Assuming an example applied to the simulation game of "Taro"
It shall be. Generally, a character with a rich personality is set up.
A set of interactive application software
Must be applicable as a human interface
Of course. This system includes an input unit 101, as shown in FIG.
Interface unit 2, information analysis unit 103, situation diagnosis unit 104,
Self-diagnosis unit 105, decision-making unit 106, action selection unit 107, operation
It comprises a generation unit 108, an output unit 109, and the like. The input unit 101 is used to input a character such as an instruction or a question.
Power information, or information about the surrounding conditions such as weather or the environment.
A device for the operator to input
And joysticks. The interface unit 102 includes an input unit 101, an output unit 109, and the like.
Interface with external devices and to which character
Recognition of the input is performed. The information analysis unit 103 is input via the interface unit 102
Analyzes the data, confirms the type of the data,
Check the contents of the data. The situation diagnosis unit 104 displays the words and
What kind of situation does the character itself respond to changes in the surrounding environment
Is determined by fuzzy pattern matching, and the action selection unit 10
7 input. The self-diagnosis unit 105 performs psychological mechanism processing and mental information
A diagnostic process is performed. Its psychological mechanism processing
Is the psychological effect of the entered words on the character
Based on the character's own state (physical / psychological) and words
Then, a decision is made by fuzzy pattern matching,
To be input to. At the same time, a mental state
Generate a minute model. In addition, the mental information diagnosis processing
The emotional state of given instructions and actions
Fuzzy putters based on their own condition (physical / psychological)
That is obtained by the collation and used as an input to the decision-making unit 106.
It is. The decision making unit 106 responds to the
Judge and decide based on Kuta's own psychological state (psychological block)
Is what you do. The action selection unit 107 checks the information input to the character and the diagnosis.
Select the action to be taken from the results of the
This is to be input to the crop generation unit 108. The action generation unit 108 receives the information input from the action selection unit 107
Of each character based on the
It generates motion information such as motion, utterance, and facial expression. So
Operation information is appropriately output via the interface unit 108.
It is converted to a force form, and the
YA). [2] Data construction processing In the system of the embodiment shown in FIG.
Various collation data necessary for executing judgment and decision processing
Is created in advance as a database. Execution of the embodiment
Before giving a detailed description of the system,
Will be described. [2-1] Schematic Configuration of Data Generation Processing Device Data (Fuzzy Character Data) Used in the Present Invention
Is created using a construction tool for that purpose. Fig. 2
Shows the configuration of the construction tool.
Environment data generation processing unit 201 and a fuzzy character
The control data generation processing unit 202 and the fuzzy character model
A data input device 204 and a display
Generate data interactively using device 205
Has become. [2-2] Fuzzy character environment data generation processing unit FIG. 3 shows the operation of fuzzy character environment data generation.
It is shown. Environmental data from the data input section
Creation of data type database and influence factor database
Performing, modifying, writing, etc. Here, the data type data includes questions, weather, and groups.
Predefined data items such as atmosphere, actions, instructions, and talk
It refers to the type of Influence factor data is set in the mood effect history.
This is the ID and comment as shown in Fig. 5.
It has become. [2-3] Fuzzy character control data generation processing unit FIG. 4 shows a fuzzy character control database generation processing unit
It is a figure which shows the operation | movement flow of a process. First, the environmental data, the impact data
Read cause data and data type data (step
401). The fuzzy character control data to be created is shown in Fig. 5.
It has a data structure like
Type ID 502, output information flag 503, influence factor ID 504, data
Data 505. The input data ID is
It is a serial number given to input data. The data of the read data type database is displayed in a table.
When displayed on the display device 205, the operator (sets data
The operator) selects a data type (step 402) and controls
The data type ID in the data is determined. Next, the operator describes output information such as actions, utterances, and facial expressions.
And set whether or not to output as a flag (step
403). Also, in the same manner as in step 402, the operator
Is selected (step 404), and the influence factor ID is determined. Then, data type corresponding data is set (step
405). The content of the data corresponding to the data type,
For example, when the data type is a question,
“Why?” Is the corresponding data. One control determined or set as described above
Data is written to a fuzzy character control database
(Step 406). Until all control data has been generated,
The generation process of each control data by step is repeated
When all control data has been created,
The process ends (step 407). [2-4] Fuzzy character model data generation processing unit FIG. 6 shows creation of a fuzzy character model database
It is a figure showing the operation flow of processing. Here, for example,
Define what kind of character Momotaro is
Fuzzy character model data
To achieve. The data structure of the fuzzy character model is No. 7.
As shown in the figure, the character ID, the value of each model,
Hiking history data, data fields such as group ID
Has become. First, a memory area for creating fuzzy character data
Is secured (step 601). An initial value is set for each use model (step 602). Refer to the influencing factor database for each usage model and
The sound history is initialized (step 603). Group database (See Figure 8 for the data structure
)) And refer to the group
Set the group ID to which the character belongs (step 60
Four). If you create a new group,
Register the loop ID in the database. Pattern data that is fuzzy character personality data
A table (see FIG. 15) is created (step 605).
Also, a filter for emphasizing individuality is set. All the above settings are completed.
The process is continued (step 606). All settings
When finished, select the character set in steps 602 to 604.
Data from the fuzzy character model database.
(Step 607). The pattern data created in step 606 is
Write to the j character personality database (step
608) [3] Detailed configuration of the embodiment FIG. 9 shows the configuration of FIG. 1 in more detail.
You. The situation diagnosis unit 902 includes a question diagnosis processing unit 903 and a group atmosphere.
Equipped with an atmosphere setting processing unit 904 and an environment change setting processing unit 905
Weather model, physical fitness model, group atmosphere model
Database 906 for damage and damage models for diagnosis
Referred to. The self-diagnosis unit 907 includes the mental information diagnosis processing unit 908 and the psychological
It has a mechanism processing unit 909, and has a mood model and
The database 910 of the healthy model is referred to. The decision making section 911 has a decision making section 912. The action selection unit 913 is an action selection processing unit 914, an utterance selection processing unit
915 and an expression selection processing unit 916 are provided. The action generator 917 includes an action generator 918, an utterance generator 919,
And a facial expression generator 920. The function of each unit will be described below. [3-1] Information Analysis Unit The information analysis unit 901 checks the type of the input control data.
Confirmation (step 1101) and data contents confirmation (step 11)
02) and passes the control data to the situation diagnosis processing unit 902.
You. [3-2] Situation Diagnosis Unit FIG. 11 is a diagram showing a flow of the situation diagnosis processing. Status
Output to the diagnostic processing unit 902 by the information analyzing unit 901
A character of the format shown in FIG. 5 for the specified character
Ractor control data is input. Also, specified characters
Data of the physical model, damage model and energy model
Data is entered. First, the data type of the input character control data
Is determined (step 1101). Data type is question, group
When it is a loop atmosphere and weather model, the situation diagnosis unit
902 and the other data types
When talking, instructing, or acting, the input character system
The control data is output to the self-diagnosis processing unit 907. If the data type is a question, use a hierarchical question
Perform question-level operations and diagnostic processing for questions
02). If the data type is group atmosphere, enter
Rewrite the group atmosphere model data accordingly (step
1103). Then, convert the group atmosphere model into a pattern ID and
To the action selection processing section. When the atmosphere gets worse
When the sighs sigh or the weather is getting worse
To say words like it's raining,
The pattern ID is output to the action selection unit (step 1104). The pattern ID is the result of pattern matching.
It is a number indicating the result. It has 5 levels from 0 to 4
0 is the worst, 4 is the highest, 2 is normal, 3 is 2 and 4
It is in the middle. Model data should take values from 0 to 100
Is divided by 25.
Thus, five levels of pattern IDs 0 to 4 are obtained. Even when the data type is weather, the above group atmosphere
The same processing as in the case of is performed. Ie weather according to the input
Rewrite model data (step 1105) and create pattern ID
(Step 1106), and outputs the result to the action selection unit 913. [3-21] Question Diagnosis Process FIG. 12 is a schematic diagram of an answer process when a question is input.
It is a figure showing an example of a flow. Player has a specific character
And input a question, the information analysis
Recognition and character ID and question ID
Enter 903. Fuzzy putter in the question diagnosis processor 903
Situation diagnosis for the question is performed by collation. Diagnostic
The utterance selection processing unit 914 selects an utterance according to the result. So
Then, the utterance generation unit 918 creates an utterance sequence, and
Is displayed. [3-211] Question level control and question diagnosis (Fig. 13) Questions are organized in four levels from level 0 to level 3
Diagnosis processing according to each question level
Is performed. The question level indicates the depth of the question. Level 0 queries the overall state of the character
Is what you do. Level 1 inquires about the partial state of the character
Or inquire about factors for Level 0 answers
Is Umono. Level 2 queries to find the cause of level 1 status
Is what you do. Level 3 queries to further determine the cause of the condition
Is Umono. Figure 13 shows the overall flow of question level control and question diagnosis.
FIG. In FIG. 13, in order to determine the question level, first,
The question ID is determined (step 1301). Question ID is "How
If not, then it also shows the overall state
It is determined whether the question is a question (step 1302). Overall quality
Questions include, for example, “Can you fight?” “How are you?”
This is a question whose answer is determined by collating the number model as a whole.
Questions other than the general status question include, for example,
One specific model, like "What about damage?"
Or a particular primary reference primary model and auxiliary secondary
This is a question whose answer is determined by the combination of models. If the question is of overall status by the above judgment,
Set 0 to the register that sets the question level (step
1303). Then, multiple model matching as described later is performed.
(Step 1304). Next, a main / sub mode as described later is applied to the matching result.
Perform an answer (level 0 answer analysis) by Dell verification (step
1305). Further, the sub model in step 1305 is regarded as the main model.
To perform primary and secondary model matching for direct answers (step
1306). In addition, analyze the model's impact history as described below.
Then, an answer is requested (step 1307). The collation result is stored in the diagnosis result table (FIG. 31). In step 1302, it is a question other than the general status
Is set, the question level register is set to 1.
(Step 1309). And the primary and secondary models for direct answers
Dell matching is performed (step 1310). And the above-mentioned steps
Steps 1306 to 1308 are performed. In step 1301, the question ID is “Why?”
When it is determined, the current question level is determined
(1312) If the question level is 1, set it in the model work.
The specified model (Fig. 36, step 3607) is the main model
The main / sub model matching as described below (for level 0 answer
(Step 1313). And step 13 above
The processing from 06 to 1308 is performed. As a result of the question level determination in step 1312, the question level
Is 2, the model set in the model work
Analyze Dell's impact history and ask for answers (step 130
7). As a result of the question level determination in step 1312, the question level
Was 3, the question frequency analysis was performed as described later.
A response is made (step 1314). [3-212] Answer corresponding to question level 0 (FIG. 14) FIG. 14 shows question level 0, that is, the overall character
Answers by multi-model matching to status queries
FIG. 14 is a diagram for explaining the processing, and is a step 1304 in FIG.
This shows the details of the processing. Inputs include a character ID specified by the player and
And question ID, target model indicating character status, character
Pattern data from the Kuta personality database (Fig. 15)
There is. Character ID is the ID of the character specified by the player
It is. The question ID is an ID of a question issued by the player. versus
The elephant model is information about the specified character.
Has a data structure as shown in FIG.
Initialized with data from model database. This
In this embodiment, the information and group
There is information indicating a loop state and information indicating an environmental state. That
The specific contents are as follows. Information that represents the state of the character itself is based on the physical
Physical strength model (physiological state), representing physical strength, physical condition
Damage model (functional state), momentum, condition
(Nori) energy model (mental state), mental state
There is a mood model (psychological state) that expresses the state. Group
Group atmosphere model (group form)
State). Good weather conditions as information indicating environmental conditions
There is a weather model that represents the state. The state of each character for multiple model matching
And the individual pattern data of the target model.
A pattern matrix expressing the characteristics is used. Fig. 15
Shows the configuration of the matrix of pattern data for matching.
It is shown. The pattern data is shown in Fig. 16 in this example.
As shown, the L (Low), M (Middle), and H (High)
It is described in symbols. Patterns, not symbols
Needless to say, it is possible to describe up to. Returning to the description of FIG. 14, the designated character and the question
The matrix of pattern data corresponding to
It is selectively read from the base (step 1401). example
If you specify Momotaro as a character, "Can you fight?"
If asked, work on the question "Can you fight?"
Pattern data of the created personality database (No. 15
(The pattern data as in the first sheet in the figure) is selected. Next, a character indicating the status of the specified character
Collating model data (Fig. 7) with the above pattern data
(Step 1402). And the best match putter
Is the worst (0) to the best (4)
Pattern ID and output the pattern ID (step
1403). Note that the pattern description and matching method
JP-A-2-98738 (Applicant Co., Ltd. Add-in Research)
Tokoro Nakamura, three inventors, "Knowledge compilation pattern type"
Inference system) can be used.
Here, in simple terms, pattern data is normalized
Center value P describing the obtained data, comparison original function ｛f1
(P), f2 (v), f3 (va) Change the characteristics of F｝ (N)
Value V for changing the characteristics of the comparison original function, the degree of warmness Va
Composed of The degree of matching is defined as the specific gravity value g.
Can be obtained by the following equation. [3-213] Answer corresponding to question level 1 Question level 1 is an inquiry about the partial state of the character
This directly includes the partial state of the character.
Questions to check () and answers corresponding to question level 0
On the other hand, a question to find the cause of "Why?"
There is. For questions that directly check the partial state of the character
Answering process (Fig. 17) Fig. 17 shows the quality of checking the partial state of the character directly
FIG. 13 is a diagram for explaining the answer process to the question, and FIG.
It shows the details of the processing of steps 1306 and 1310 of FIG. Inputs are character ID, question ID, primary model, secondary model,
And pattern data of the main and sub models. In this case, the main model is determined by the question ID.
The sub model is fixed to the main model as shown in the following table.
Assigned in a way. Primary model Secondary model Physical strength Genki Group atmosphere Mood Group atmosphere Environment Mood Damage Energy Group atmosphere Mood Individual pattern data corresponding to character ID and question ID
Selectively from the sex database (step 170)
1). For example, if you specify Momotaro as a character,
If you ask "How is your physical strength?"
The physical model has been specified as the model,
As for the sub model, a fine model is designated from the table. And
A pattern model that uses Momotaro's physical strength and spirit as the main and sub models
The data matrix is read. FIG. 18 (a) shows the main and sub
9 shows an example of model pattern data. putter
Individual models with weight filters for primary and secondary models
Is emphasized, and individuality is described by the pattern description.
You. Specifically, the filter ratio is basically 7 to 3,
The ratio has been changed on a fractional basis. Then, the main / sub model representing the state of Momotaro is
Check with the turn data (step 1702). And most
Patterns with a high degree of matching are worst (0) to best (4)
Determine whether the pattern is in a misaligned state, and
The ID is output (step 1703). Answer process for the question "Why?"
Figure) Figure 19 shows how responses to question level 0
Shows the answer process when asking the question
13 shows details of step 1310 in FIG. Corresponding pattern used in level 0 question in Figure 13
A matrix of data is obtained (step 1901). next
Model with the highest contribution to level 0 questions
(Step 1902). Overall match level 0
Of each model corresponding to the highest “state” of
The value multiplied by the value of the weight filter is defined as the contribution ratio. The model with the highest contribution rate is set as the main model,
A predetermined model corresponding to the model is defined as a sub model.
Pattern data of main and sub models from the personality database
read out. And the current target for the pattern data
(1903). Collation
As a result, the pattern ID of the “state” having the highest matching degree is obtained.
(Step 1904). For example, as shown on the left side of FIG.
In response to the question "Can you fight?"
"No!" And ask "why?"
Suppose. In the answer process in Fig. 19, the answer is "absolutely unpleasant!"
To find the model that has the highest contribution rate. For example,
If the model with the higher rate is the physical fitness model,
The model is the main model, and the fine model is the sub model.
To the corresponding pattern data. As a result,
When the state is the worst match,
"I have no physical strength!" [3-214] Answer corresponding to question level 2 (FIG. 20) FIG. 20 shows the answer for question level 2 or question level 1.
Query to find the cause of the "state" of the model in
FIG. 13 is a diagram showing a response process based on the influence history, and FIG.
It shows the details of step 1307. Influence of target model from character model database
The history information is taken in (step 2001). And targeted
Select a model history. Impact history is + factor,-required
Each set of factors is represented by an ID (step
2002). At the question level 1
Flag indicating whether the answer is good or bad
Of the impact history data by referring to the flag.
Retrieves factor data and outputs factor ID to self-diagnosis unit 105
(Step 2003). For example, in question level 1 processing,
"" At no point, and "
Search the history of the effects of the physical fitness model
Then, the + factor is ID "5", which represents "rest", and the-factor is
It is assumed that the ID is “10” representing “fought”. The flag
In this case indicates that it is on the-side, so-
Is output to the action selection unit 913, and the corresponding
Display the answer "Because I fought"
You. [3-215] Answer corresponding to question level 3 (Fig. 20) Fig. 21 shows the answer for question level 3, ie, question level 2
To further investigate the cause of the "state" of the model in
Is a diagram showing answer processing by question frequency analysis for
13 shows details of step 1314 in FIG. Answer processing for this question level 3 is a psychological mechanism
This is performed by the system processing unit 909 and the decision processing unit 912.
You. Fixedly prepared in the psychological mechanism processing unit 909
From the adverse effect data and the current model state
Influences the mood model by dynamic mechanisms (step
Step 2101). Fig. 26 shows the psychological mechanism.
The details will be described later. Next, in the decision processing unit 912, the
Determine the answer level to the question (step 210)
2). The action selection unit 913 uses the response level as a pattern ID.
Output to Fig. 29 shows the psychological block.
Details will be described later with reference to FIG. 30. Method of creating impact history For questions (Figure 22) Figure 22 shows the process of creating an impact history for questions
It is something. Count the number of consecutive questions of the same question,
If the same question is repeated three times, the mood model-the influential factors
Set the factor ID to the side. For example,-
Is set to ID "10" indicating " Talking (Figure 23) Figure 23 shows the process of creating an impact history for talking.
It is something. State (physical strength, mood ...), spoken words (giving pressure
Stuff that gives a sense of openness)
Mood model using pattern data by dynamic mechanism processing
Dell pattern matching is performed (step 2301). Matching result
It is determined whether the fruit feels good or not (step 2302).
Was in a good state, the positive side of the mood model
Set the factor ID to (2303).
The factor ID is set to the factor side (2304). For example, with the state
Input a conversational language such as "as expected"
In good mood, based on psychological mechanisms
If there was, “complained” to the positive influence side of the mood model
Set an ID that represents “ta”, and conversely,
You are "weird" on the negative side of the mood model
Set the factor ID that indicates In the case of instruction / action (Fig. 24) The state of the character corresponding to the instruction / action
Is diagnosed by mental information diagnosis (pattern matching).
Step 2401). And find a pattern ID with a high degree of matching
(Step 2402). The obtained pattern ID is the character's own
Determine whether it is good for yourself (2403) and determine that it is good
If it is, +
An ID is set on the influence factor side (step 2404). Good
If the judgment is not true, the model affected by the instruction or action
The ID is set on the negative side (step 2405). example
For example, when the instruction of "fight" is
If the session ID is 0 (worst case),
The corresponding mood model-the influencing factor side
Is set to a factor ID = 10 representing "weird instruction". [3-3] Self-diagnosis unit [3-31] Psychological mechanism processing (FIG. 26) FIG. 26 is a diagram for explaining psychological mechanism processing
Yes, an example is shown for conversation (speaking). What
Contact, the whole answer process for talking (same for instructions and actions)
The typical flow is shown in FIG. Story selected by the player
The psychological mechanism using the mood model and the spirit model
Perform self-diagnosis by nysm processing and further psychological block
The response level is determined by the
Select and make an utterance and display it on the screen. Character personality data of characters targeted for conversation
That is, the pattern data matrix is fixed to characters
It is selectively read from the database (step 2601). The current self indicated by the model data of the target character
Enter the state of the body and the entered conversation language (front, back)
Good mood patterns in the character personality data
Are collated (step 2602). No. for conversational languages
27 As shown in Figure 27, the meaning of the table and the meaning of the back are expressed as numerical values
Can be predetermined. For example, "I'm alright!"
When the conversation language is input, the front is 80 languages and the back is 10 languages.
Treated. An example of the matching pattern is shown in FIG. like this
In the present embodiment, pattern data (individuality) relating to mood is used.
In the next four are prepared, characters for verification
The combinations of the data models are as follows. Personality character model When encouraged Physical strength, mood, damage When praised Physical strength, mood, damage Physically open feeling Physical strength, energy, damage When pressure is applied Energy, mood, group Atmosphere Condition (physical strength, mood,
Damage) and the degree of matching between the pattern data. Ma
At the same time, depending on the calculated degree of matching,
Increase or decrease the value of the mood situation model against the criteria.
A good standard situation model is 4
Four corresponding to each of the character personality data
It is prepared. Next, in the same way, the bad
Performs matching with mood patterns and calculates by matching
Bad mood situation model depending on the degree of matching
The status model value is updated (step 2602). The bad group
The quasi-situation model also has four characters related to the aforementioned mood.
There are four corresponding to each of the personality data
You. Good mood situation model and bad mood situation model obtained
Based on a mood situation model that includes both values of
Then, a mood model is created (step 2604). Sand
If you are encouraged or praised,
If you are given a physical sense of openness,
Of the character personality data in the four cases
Four good mood situation models provided for each
And the values of the four bad mood situation models
Generate That is, the mood model is calculated by the following equation
I do. Mood model = 50 + (Good mood situation model total) / 4-
(Bad mood situation model total) / 4 Next, the long-term mood is calculated based on the mood model value obtained above.
Of short-term trends based on matching results
Step 2605). The long-term trend is that the mood model value is 0
49 is a bad tendency and 50-100 is a good tendency. Asked
The mood trends are based on the long-term mood
Field. Short-term trends are determined in steps 2602 and 2603.
The side with the higher degree of matching of fruits is a short-term tendency of mood. Judge whether the short-term tendency found is good
(Step 2606). The short-term trend we sought was good
When the mood effect history (+ side) is
An ID corresponding to the taste is set (step 2607). Also,
If the short-term tendency was bad, the mood effect history
Set the ID corresponding to the input global meaning on the (-) side
(Step 2608). The pattern ID is obtained from the value of the mood model. That is,
Pattern ID = Mood model value / 25 (Step 2
609). Mainly for each model of health, energy and damage
Check the sub model and store it in the diagnosis result table (step
2610). [3-32] Mental information diagnosis processing FIG. 28 shows the flow of mental information diagnosis processing.
Mental information diagnosis processing is performed as self-diagnosis of behavior and instructions.
Processing. Character personality data (pattern data) corresponding to the instruction
Data is selectively read from the character personality database.
(Step 2801). Character model that represents the current character situation
The data is collated with the personality data (step 2803). As a result of the comparison, the password corresponding to the “state”
A turn ID is obtained (step 2804). Check the value of the pattern ID (step 2805), and check the pattern ID.
When = 0, the processing proceeds to the psychological mechanism processing described in FIG.
Give more adverse data and reflect it in the mood model
(Step 2806). When the pattern ID is 4, the psycho-mechanical
Improve mood model by giving good influence data by nism processing
It is reflected (step 2807). Also, if the value of the pattern ID is
When the value is an intermediate value, that is, when the pattern ID is 1, 2, or 3,
It is not reflected in the minute model. Character personality data selected in response to instructions or actions
Data (pattern data) to determine whether it is a multiple model
(Step 2808). If the result of the determination is that there are multiple models,
Dell matching is performed (step 2809). This model
Is the main / sub model matching of the above question (level 0 answer analysis)
In the same way as described above, a model having a high contribution rate is searched and collated. If the result of the above determination is that the model is a single model,
Performs Dell matching (direct) and replaces the sub model with the main model
The main / sub model matching is performed (step 2810). This
Model matching for instructions or actions (direct)
Is the same process as the main / sub model matching (direct)
is there. Each collation result is stored in the diagnosis result table, and step 28
The mood model data updated in 06,2807 is the mood model memory
Store in area. [3-4] Decision Making Unit FIG. 29 is a flowchart for explaining the operation of the decision making process.
-It is a figure. Decisions are made to take into account the psychological state of the character
This is the process to be performed. First, the input is an instruction or action
Or it is a talk (step 290).
1). If it is an instruction or action, diagnose the instruction or action
The result is obtained from the diagnosis result table (step 2902). Using the mood model and the diagnosis results as input,
Pattern matching is performed (step 2903). Psychological block
Is the one stored in the character personality database
And the pattern data of the mood data as shown in Fig. 30.
Character by the combination of
Pattern data matrices representing various mental states of Kuta
It is a box. The pattern ID with the highest match as a result of pattern matching
(PID) is stored in the diagnosis result table (step 290).
Four). For example, if the instruction is “fight”,
If the multiple model matching pattern ID in the result table is 2 (normal)
And the pattern ID of the mood model was 3.
Then, the degree of matching with the psychological block is “good” (pass
Since the turn ID is the highest at 3), the diagnostic result table shows
Update the multiple model matching pattern ID to 3. [3-5] Action Selection Unit FIG. 32 is a flowchart for explaining the processing operation of the action selection unit.
-It is a figure. The input to the action selection unit 913 is character control data,
And diagnosis generated by the situation diagnosis unit 902 or the decision making unit 911
Each data of the disconnection result table is used for processing. First, check the output information flag of the character control data.
To determine whether the utterance flag is on or off
(Step 3201). When the output information flag is on,
If you are instructed to do so,
Input and collate with the utterance selection pattern (step
3202). The utterance selection pattern has a matrix as shown in Fig. 33.
Rix, character personality database
It is one of those stored in. Diagnosis result table
Various combinations of pattern data of the same items
The content of the story is decided. The utterance content is
11 consequent remarks like “absolutely no” (pattern ID
0 to 10), issues related to models such as “no physical strength”
20 stories (pattern ID = 11-30), like "fought"
20 utterances about the history (pattern ID = 31-50),
10 mood-like utterances such as "What?"
= 51-60) are prepared for each. Such utterance selection
Diagnosis result table for each pattern data of selected pattern
Of each item by comparing with the data of each corresponding element of
Calculate the degree of match, and count each line by counting in the row direction
Is determined. Then, the pattern IDs of the top five utterances with the highest matching degree are selected.
And store it in the utterance data table (step 320).
3). The data format of the utterance data table is shown in Fig. 37.
The number of data and the pattern ID of the selected utterance
Field. In step 3201, check the output information flag of the control data.
If the utterance flag is determined to be off, the utterance data
0 is set to the number of data in the data table (step 320).
Four). Next, behaviors and facial expressions were almost the same as in conversation.
The same processing is performed. In other words, regarding the behavior, the character control data
Check the output information flag and check if the action flag is on or off
Is determined (step 3205), and the action output information flag is determined.
When the diagnosis is on, the diagnosis result table is
It is compared with the selection pattern (step 3206). Action selection pattern
The matrix consists of a matrix as shown in Figure 34.
You. And the pattern ID of the top one
Select and store in the action data table shown in Fig. 38
(Step 3207). In step 3205, the action flag is turned off.
If it is determined that there is, data in the behavior data table
The number is set to 0 (step 3208). Similarly, in the case of outputting a facial expression, the facial expression selection
Check the diagnosis result table for the turn (step 321)
0), the pattern ID with the highest matching degree is shown in Fig. 39
(Step 3211). table
If you do not want to output the information,
The number of data 0 is set (step 3211). [3-6] Action generation unit The action generation unit 917 generates the utterance and line generated by the action selection unit 913.
Based on each data table
Generate crops. A place for generating utterance data based on FIG.
The case will be described. Check the number of data in the utterance data table (step 360
1) If there is data, enter the utterance data table
And collate with the utterance generation pattern as shown in Fig. 40.
(Step 3602). Find the pattern ID with the highest matching degree as a result of the matching
Step 3603). Corresponds to the pattern ID and enters the data in the utterance data table.
It is replaced and deleted (step 3604). That is, utterance
Copy the data in the data table to the pattern with the highest match
Replace with the utterance generation method corresponding to the ID. From the utterance pattern ID in the utterance data table, the final
The utterance level is obtained (step 3605). Speech level is 0
It is checked whether it is or 1 (step 3606), and 0 or 1
Is high, the match level is high from the utterance data table.
(The side near the top in the table)
First, set the model ID in the model work (step 360
7). The pattern ID set in the model work is shown in Fig. 13.
Used to specify the main model at the time of main / sub model matching. Ma
Set the utterance level +1 to the question level (step
3608). If the utterance level is neither 0 nor 1, quality
Set the question level to 3. These question levels are shown in Figure 13.
It is referred to in the question level determination process.

【The invention's effect】

According to the present invention, since the personality of a character is set by fuzzy pattern data of a plurality of models, character characterization and action patterns of the character can be set to complex and various types. In comparison, the function of the human interface of the program to which the present invention is applied can be significantly improved. When applied to games, it becomes more interesting, when applied to mental treatment, the content of treatment can be deeper and more accurate, and when applied to CAI, more appropriate educational content according to the child's situation Can be controlled, and the function can be improved. Further, since the setting, change, and correction of the character's individuality can be easily performed by setting and changing the fuzzy pattern data, the number of steps for developing the program can be greatly reduced. Conventionally, the program capacity increased in proportion to the complexity and sophistication. However, according to the present invention, since the fuzzy pattern data of a plurality of models is used as described above, the program capacity becomes extremely small, and the memory area is limited. Even if it runs on a computer, it is possible to develop a flexible story with complex characters. Further, according to the present invention, since the character can be provided with a psychological model, the character can exhibit a more human response.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention. FIG. 2 is a diagram showing a configuration of a construction tool for creating data (fuzzy character data) used in the present invention. FIG. 3 is a diagram showing an operation of generating fuzzy character environment data. FIG. 4 is a diagram showing an operation flow of a fuzzy character control database creation process. FIG. 5 is a diagram showing a data structure of fuzzy character control data to be created. FIG. 6 is a diagram showing an operation flow of a fuzzy character model database creation process. FIG. 7 is a diagram showing a data structure of a fuzzy character model. FIG. 8 is a diagram showing a data structure of group data. FIG. 9 is a diagram showing details of the configuration of the embodiment of the present invention. FIG. 10 is a flowchart showing the operation of the information analysis process. FIG. 11 is a flowchart showing the operation of the situation diagnosis processing. FIG. 12 is a diagram showing an example of a schematic flow of answer processing when a question is input. FIG. 13 is a flowchart showing the operation of the question level control and the question diagnosis processing. FIG. 14 is a diagram for explaining a process of answering by inquiry of a plurality of models in response to an inquiry about question level 0, that is, an overall state of a character. FIG. 15 is a diagram showing a configuration of a matrix of pattern data for matching. FIG. 16 is a diagram showing an example of pattern data. FIG. 17 is a diagram for explaining the answer process to the question for directly confirming the partial state of the character. FIGS. 18 (a) and 18 (b) show examples of pattern data matrices, respectively. Figure 19 shows why the answer to question level 0
Figure 20 shows the answer process when asking the question "?". FIG. 20 shows a query to find the cause of the "state" of the model in the answer at question level 2, ie, question level 1.
It is a figure which shows the response process by an influence history. FIG. 21 is a diagram showing answer processing by question frequency analysis for an inquiry for further searching for the cause of the “state” of the model in the answer at question level 3, ie, question level 2. FIG. 22 shows a process of creating an influence history in the case of a question. FIG. 23 is a diagram showing processing for creating an influence history in the case of speech. FIG. 24 is a diagram showing a process of creating an influence history in the case of an instruction / action. FIG. 25 is a diagram showing an overall flow of a reply process for a talk (same for an instruction and an action). FIG. 26 is a diagram for explaining the psychological mechanism processing. FIG. 27 is a diagram for explaining the front and back meaning expression of the conversation language. FIG. 28 is a diagram showing a flow of a mental information diagnosis process. FIG. 29 is a flowchart for explaining the operation of the decision making process. FIG. 30 is a diagram showing an example of a psychological block pattern. FIG. 31 is a diagram showing an example of a diagnosis result table. FIG. 32 is a flowchart for explaining the processing operation of the action selecting unit. FIG. 33 is a diagram showing an example of an utterance selection pattern. FIG. 34 is a diagram showing an example of an action selection pattern. FIG. 35 is a diagram showing an example of an expression selection pattern. FIG. 36 is a flowchart showing a motion generation process (for utterance generation). FIG. 37 is a diagram showing a configuration example of an utterance data table. FIG. 38 is a diagram showing a configuration example of an action data table. FIG. 39 is a diagram showing a configuration example of an utterance data table. FIG. 40 is a diagram showing a configuration example of an utterance generation pattern. 101 input unit, 102 interface unit, 103 information analysis unit, 104 situation diagnosis unit, 105 self-diagnosis unit, 106 decision-making unit, 107 action selection unit, 108 Motion generation unit, 109 ... Output unit.

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Shigeru Oyamada 1-15-8 Jinnan, Shibuya-ku, Tokyo Kennaka Building 4th floor Inside Add-in Research Institute, Inc. (56) References JP-A-2-159625 (JP, A ) "Computer software" Vol. 4, No. 2 (1987-4) p. 24-38 “Proceedings of the 1989 Annual Meeting of the Japanese Society for Artificial Intelligence (3rd)”, p. 617-620 (58) Field surveyed (Int. Cl. ⁷ , DB name) G06F 9/44

Claims (5)

(57) [Claims] An input unit capable of inputting instructions or questions from an operator in the form of words, a target model having a plurality of models including a psychological model, and fuzzy pattern data corresponding to each model to determine the personality of the character. Diagnosis for diagnosing the psychological situation of the character itself by comparing the database storing the expressed fuzzy character personality data with the information of the instruction or question in words input from the input unit and the fuzzy character personality data fuzzy pattern data A fuzzy pattern human, comprising: an operation selecting unit that selects an action to be performed based on a diagnosis result of the diagnosis unit; and an operation generating unit that generates an operation corresponding to the selected action. Interface system. And a diagnostic section for determining a state of the character itself in response to a word input to the character or a change in surrounding environment by comparing the character with fuzzy pattern data, and outputting a diagnostic result. The fuzzy pattern type human interface system according to claim 1, further comprising: 3. The diagnostic unit obtains a psychological effect of a word input from an input unit on a character by comparing the psychological effect with a fuzzy pattern data based on the character's own state and words, and outputs a diagnostic result. The fuzzy pattern type human interface system according to claim 1, further comprising a psychological mechanism processing unit for generating a mood model representing a psychological state. 4. A mental unit which obtains a mental state with respect to an input instruction and an action of the character by comparing it with fuzzy pattern data based on the state of the character itself and outputs a diagnostic result. The fuzzy pattern type human interface system according to claim 1, further comprising an information diagnosis processing unit. 5. A decision processing unit which receives the mood model and the diagnosis result as input, compares it with fuzzy character pattern data representing various mental states of the character, and makes a decision on a matter to be determined. The fuzzy pattern type human interface system according to claim 3 or 4, wherein: