Information processing device, information processing method, and program

Abstract

To make a plurality of thoughts associated with a plurality of personas, and treat them in a multiple manner.SOLUTION: An information processor marks a first element gaining topicality in a first rule tree as a value-acquired element, when a second element having an alias the same as and similar to the marked first element exists, marks the second element as a value-acquired element, determines whether all the elements existing between a top node and a target element are marked, for each of the first rule tree associated with the marked first element and a second rule tree associated with the marked second element, when all the elements are not marked, and selects an element specified on the basis of a predetermined reference, among the elements which are not marked in the first rule tree and the elements which are not marked in the second rule tree, as next topicality.SELECTED DRAWING: Figure 7

Description

[Technical field]

[0001] The present invention relates to an information processing device, an information processing method, and a program. More specifically, the present invention relates to an information processing device, an information processing method, and a program that associates multiple thoughts with multiple personas (rule trees), controls them so that they can be handled in a multiple manner, and provides output results in more natural language through a conversion function such as a generative AI. The present invention also relates to an information processing device, an information processing method, and a program that adds a combination of elements and element values ​​derived from new input data to case data, brushes up the persona, and generates a persona based on input data that is associated with an end user. [Background technology]

[0002] An expert system in the field of artificial intelligence technology (Patent Document 1) generates a decision tree by registering elements (condition attributes) and element values ​​(condition attribute values) for a specific theme, and further registering a target element (conclusion attribute) and a target element value (conclusion attribute value). The expert system can provide consulting to a user on a specific theme by using the decision tree.

[0003] A personaization system (Patent Document 2) is also known as a technique for generating a decision tree. The personaization system is equipped with a case learning / machine learning tool. This tool expresses cases based on input data as a set of elements (condition attributes) and target elements (conclusion attributes), and also learns rules for estimating target element values ​​(conclusion attribute values) from element values ​​(condition attribute values) observed based on cases to generate a decision tree (hereinafter referred to as a "rule tree"). This tool outputs a target value by automatically tracing the rule tree based on the way of responding (element values) to preset question items (elements) (auto-consulting). Patent Document 2 discloses a technique for moving between multiple rule trees, which is performed by setting the address of the rule tree to be called to an arbitrary element as key information.

[0004] In recent years, a technology called generative AI has emerged in the field of artificial intelligence. Although generative AI is still in the early stages of development, it has the ability to generate and output new text, images, music, etc. based on input information and previously learned information. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 5572615 [Patent Document 2] Patent No. 7216449 Summary of the Invention [Problem to be solved by the invention]

[0006] The problems of modern society are complex. At first glance, the problems people face may seem like they can be categorized into a specific theme, but in reality, problems may arise from a mixture of various factors and a combination of various causes. When people talk to each other about a certain theme or problem, the topic may change to another theme or problem when a certain keyword comes up. In such cases, consulting cannot be done by experts in a specific field alone, and cooperation from experts in other fields is required.

[0007] In the conventional technology, a rule tree is generated in advance in order to provide consulting to a user on a specific theme or problem. Even if a rule tree generated in this way is used, it is difficult to provide appropriate consulting on complex themes or problems. In order to be able to handle multiple themes and problems, it is necessary to generate a rule tree that can handle multiple thoughts. However, in order to provide consulting to a user on various problems, it is necessary to generate a huge rule tree that combines multiple thoughts into one.

[0008] FIG. 1 is a diagram explaining the process of generating one huge rule tree in the prior art. The huge rule tree is generated based on various elements and examples so that consulting can be performed for various problems. Although the huge rule tree has the advantage of being able to handle multiple thoughts, (1) It takes a huge amount of time to complete the process because it requires a large number of examples to generate. (2) Considering the possibility of branching out into various themes and issues, the amount of information (elements) required for one case is large, and (3) If it is later discovered that a necessary element is missing or has been added, it will be necessary to add the element to all cases. There are disadvantages such as:

[0009] When a topic shifts from a specific theme or problem to another theme or problem, it is necessary to set the address of the rule tree to be called as key information to an arbitrary element, as disclosed in Patent Document 2. FIG. 2 is a diagram explaining the concept of setting the address of the rule tree to be called as key information to an arbitrary element. In FIG. 2, the address of the called rule tree is set to the shaded element. With such conventional technology, it is difficult to decide which address to set to which element, and in some cases, there is a risk that the rule tree will become rigid.

[0010] In conventional consulting using rule trees, the target element value of the tree traversed as a result of user dialogue was output as is, which sometimes resulted in unnatural linguistic output.

[0011] Conventionally, it was possible to generate a rule tree from cases, but the cases themselves had to be prepared by the rule tree generator. It was a heavy burden for the generator to prepare appropriate case data including elements and values, and it was difficult for the generator to determine how to set elements and how to ensure consistency with existing personas.

[0012] For this reason, there is a demand for information processing that can easily construct multiple thoughts using multiple rule trees and automatically select a rule tree for a desired theme or problem.In addition, there is a demand for information processing that can convert the contents of target element values ​​of one or multiple rule trees obtained as a result of user dialogue into more natural language expressions.Furthermore, there is a demand for information processing that can easily generate case data that is consistent with the elements and element values ​​of existing personas based on new input data.

[0013] The present invention has been made to solve such problems, and aims to provide an information processing device, information processing method, and program that associate multiple thoughts with multiple personas (rule trees) and control them so that they can be handled in a multiplexed manner.

[0014] The present invention aims to provide an information processing device and an information processing method that, when target element values ​​of multiple rule trees are obtained as a result of user interaction, converts the content into a more natural language expression, or generates a new natural language expression based on the content.

[0015] The present invention aims to provide an information processing device and an information processing method that add combinations of elements and their values ​​that can be derived from new input data prepared by a user as new case data in a form that is consistent with combinations of elements and element values ​​of existing personas. [Means for solving the problem]

[0016] In order to solve the above problems, the information processing device according to the present invention comprises: Mark the first element being discussed in the first rule tree as having a value; performing a query to determine whether there is a second element having the same or similar alias as the first element marked; If the second element is present, marking the second element as having been valued; For each of a first rule tree associated with the first marked element and a second rule tree associated with the second marked element, determining whether all elements between a top node and a target element are marked; On the condition that all elements are unmarked, an element identified based on a predetermined criterion from among the unmarked elements of the first rule tree and the unmarked elements of the second rule tree is selected as the next topic. It is configured as follows. Effect of the Invention

[0017] According to the present invention, multiple thoughts can be associated with multiple personas (rule trees) and controlled to handle them in a multi-layered manner, making it possible to provide general-purpose consulting for a variety of themes and problems. Since the present invention has fewer elements for each persona than the conventional technology, fewer cases are required and the amount of information per case is smaller. Even when adding a required element after the fact, the present invention requires only updating the associated persona, making the scope of the control target smaller. Also, since the present invention has relatively smaller personas than the conventional technology, it is easier to check and maintain the rules.

[0018] According to the present invention, the target element values ​​of multiple rule trees obtained as a result of user dialogue can be converted into more natural language expressions, or new more natural language expressions can be generated and output, thereby enabling the user to dialogue with the persona using more natural language expressions.

[0019] According to the present invention, even during a service using a persona (rule tree), new case data can be added and the rule tree can be updated based on the added case data, allowing the persona to grow and localization to suit the region in which the service is used can be performed.Furthermore, by preparing case data for each end user, each rule tree can grow into a tailor-made rule tree for the end user. [Brief description of the drawings]

[0020] A more detailed understanding of the embodiments disclosed herein can be had from the following description, taken in conjunction with the accompanying drawings, in which: [Figure 1] FIG. 1 is a diagram illustrating a process for generating a large rule tree in the prior art. [Diagram 2] FIG. 13 is a diagram for explaining the concept of setting the address of a rule tree to be called as key information in an arbitrary element. [Diagram 3] 1 is a diagram for explaining the configuration of an entire system including an information processing device 10, a user terminal 11, and a designer terminal 12 according to the present invention. [Figure 4] 10 is a diagram showing an example of elements and case information stored in the case data information 106. FIG. [Diagram 5] FIG. 2 is a diagram showing an example of a persona (rule tree) generated by the information processing device 10 based on case data. [Figure 6] 1 is a diagram illustrating the concept of a persona (rule tree) generated by information processing device 10. FIG. [Figure 7] FIG. 13 is a flow diagram illustrating a process for making recommendations for a user on a specific topic using multiple personas. [Figure 8] This is a diagram explaining an example of "interaction between persona and user" and information processing performed on a group of personas. [Figure 9] FIG. 13 is a flow diagram illustrating the process of adding new personas and modifying only relevant personas. [Figure 10] This is a diagram illustrating the process of configuring existing personas into new personas. [Figure 11] 13 is a diagram illustrating a process flow for converting the results of a user dialogue using multiple personas into natural Japanese and outputting the results. [Figure 12] This is a diagram explaining the concept of multiple personas generated based on case data and elements that were not included in those personas. [Figure 13]13 is a diagram for explaining the concept of processing in which the information processing device 10 adds a recognition result as case data based on the original data of a new case and combinations of all elements and element values ​​acquired by a discovery acquisition request (ALL). FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. The same reference numerals in the drawings represent the same elements, and duplicated explanations will be omitted. The examples given in this specification are merely examples, and are not intended to be limiting in any way.

[0022] (System Configuration) Fig. 3 is a diagram explaining the configuration of the entire system including an information processing device 10, a user terminal 11, and a designer terminal 12 according to the present invention. The information processing device 10, the user terminal 11, and the designer terminal 12 are connected so as to be able to communicate with each other via a network 13. Although only one user terminal 11 and one designer terminal 12 are shown in Fig. 3, there may be a plurality of these.

[0023] 3, like a general computer, the information processing device 10 includes a control unit 101, a main memory unit 102, an auxiliary memory unit 103, an IF unit 104, and an output unit 105, which are interconnected by a bus 120 or the like. The information processing device 10 includes case data information 106 and rule tree information 107 in the form of a file / database or the like.

[0024] The control unit 101 is also called a central processing unit (CPU), and controls each component of the information processing device 10 and performs data calculations, and also reads out various programs stored in the auxiliary storage unit 103 into the main storage unit 102 and executes them. The main storage unit 102 is also called a main memory, and stores various received data, computer-executable instructions, data after calculation processing based on the instructions, etc. The auxiliary storage unit 103 is a storage device typified by a hard disk drive (HDD) and the like, and stores data and programs for the long term.

[0025] 3, an embodiment in which the control unit 101, the main memory unit 102, and the auxiliary memory unit 103 are provided inside the same computer will be described, but as another embodiment, the information processing device 10 can be configured to realize parallel distributed processing by a plurality of computers by using a plurality of control units 101, main memory units 102, and auxiliary memory units 103. As another embodiment, it is also possible to provide an embodiment in which a plurality of servers for the information processing device 10 are installed, and a single auxiliary memory unit 103 is shared by the plurality of servers.

[0026] The IF unit 104 plays the role of an interface (IF) when transmitting and receiving data to and from other systems and devices, and provides an interface for accepting various commands and input data (various masters, tables, etc.) from a system operator. The output unit 105 provides a display screen for displaying the processed data and a printing means for printing the data.

[0027] The case data information 106 stores elements and case information that are the basis for generating a persona (rule tree). FIG. 4 shows an example of elements and case information stored in the case data information 106. As shown in FIG. 4, the case data includes an appropriate amount of elements and case information corresponding to the elements, which is less than that of the conventional technology, and the number of cases for generating each persona (rule tree) can be configured to be less than that of the conventional technology. The information processing device 10 generates case data including elements and case information based on the recognition result of the natural language recognition unit for new input data, or acquires case data from the designer terminal 12, and adds it to the case data information 106.

[0028] The rule tree information 107 stores information on personas (rule trees). Fig. 5 shows an example of a persona (rule tree) generated by the information processing device 10 based on case data. As shown in Fig. 5, the rule tree of the persona is relatively smaller than that of the conventional technology, making it easier to check and maintain the contents.

[0029] FIG. 6 is a diagram for explaining the concept of a persona (rule tree) generated by the information processing device 10. The rule tree has a combination of a plurality of elements 111 and element values ​​112. Any element and element value in the rule tree may be a target element and a target element value, and for example, the element value 112 at the bottom of the tree may be a target element value (i.e., the conclusion of the persona). The number of target elements in the rule tree may vary depending on the contents of the elements. In this specification, the elements 111 and the element values ​​112 may be collectively referred to as elements 100.

[0030] The element 100 may include at least the name of the event that generated the rule, the element type, the element name, the element value, the information entropy value, and an alias. The name of the event that generated the rule indicates the case (for example, "health", "body make", "cost", "food selection", "physical condition", etc. described later) from which the rule tree was generated, or the personified subject. The element type is a flag indicating whether the type of the element is a condition or a conclusion. The element name is a name indicating the contents of the condition or conclusion. The element value indicates the attribute value of the condition or conclusion. The information entropy value indicates the entropy value of the element calculated by the information processing device 10 when the rule tree was generated. The entropy value can indicate how much information the element has (also called "information value"). A high entropy value indicates that the element contains relatively more information, i.e., the information value is relatively high, while a low entropy value indicates that the element contains relatively less information, i.e., the information value is relatively low. An alias can indicate a value associated with the same meaning when a related element with the same meaning exists in the same or another persona, i.e., a synonym. An alias can have multiple synonyms.

[0031] 6, the information processing device 10 can set the name of the event that generated the rule for each element (also called a "node"), the element name, the element value, the information entropy value, and an alias. The information processing device 10 can store this set information in the rule tree information 107.

[0032] The user terminal 11 can access the information processing device 10 to call one or more personas and use those rule trees to interact with the personas, that is, to receive consulting.

[0033] The designer terminal 12 can register elements from which a persona is to be generated and case information (case data information) in the information processing device 10 via an input screen or the like provided by the information processing device 10.

[0034] An embodiment of the invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular jobs or implement particular abstract data types. An embodiment of the invention may also be implemented in a distributed computing environment where jobs are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media.

[0035] (Processing of recommendations using multiple personas) 7 is a flow diagram for explaining an example of a process for making recommendations on a specific topic of a user using multiple personas. In the explanation of this processing flow, as an example, an example of making food recommendations using multiple personas is given, and a process using the entropy value of a node is explained. The rule tree information 107 stores personas for "health", "body make", "cost", and "food selection", and each persona includes at least the following elements, but is not limited to these. "Health" persona: calories, various nutrients, etc. "Body Make" Persona: Calories, various nutrients, muscle mass, target weight, etc. "Cost" persona: food prices, contents of the fridge, etc. "Meal Choice" persona: What you ate recently, other persona targets, etc.

[0036] In S701, the information processing device 10 reads a persona associated with a topic (e.g., a "health" persona) from the rule tree information 107, and acquires an entropy value of each node of the rule tree. The topic may be received from the user terminal 11, or may be a topic that has become a candidate through the processing of S709 described below. The information processing device 10 identifies a node having the highest entropy value among the acquired entropy values ​​as a "high priority element," and executes a program associated with the identified element. For example, if the node having the highest entropy value is the top node in the rule tree, the information processing device 10 executes a program associated with the top node.

[0037] Figure 8 is a diagram explaining an example of "interaction between persona and user" executed in this processing flow and information processing performed on a group of personas. In the example of Figure 8, the top node of the "health" persona is identified and marked by the processing of S701.

[0038] Here, the entropy value of a persona will be explained. Each node of a persona (rule tree) is associated with an entropy value calculated by the information processing device 10. The entropy value indicates the amount of information (information value) that the node has, and the higher the entropy value, the greater the amount of information. In this embodiment, a node with a high entropy value, i.e., a node containing a lot of information, is considered to be a "high priority element."

[0039] In one embodiment of the present invention, the information processing device 10 can be configured to control the element selection method by changing the entropy value of each node using parameters and weighting based on an instruction from the designer terminal 12. In the present invention, the information processing device 10 can set information that controls the element selection method via a setting command from the designer terminal 12. That is, the information processing device 10 can be configured to select elements based on information associated with elements other than the above-mentioned "entropy value."

[0040] In S702, when the information processing device 10 executes a program associated with the identified element, it displays a comment associated with the element on the user terminal 11 (arrow "A" in FIG. 8). As shown in "A" in FIG. 8, the information processing device 10 can display a comment such as "How many calories do you consume per day?" on the display of the user terminal 11.

[0041] In S703, the information processing device 10 acquires the value of the user's answer from the user terminal 11 (arrow "B" in FIG. 8), and performs processing to mark the combination of the identified element and the acquired value as having acquired the value on the node of the persona. Through this processing, the top node of the "health" persona and the value associated therewith may be marked as having acquired the value. Note that, as long as it is clear that the value has been acquired, it is also possible to indicate that the value has been acquired by setting any value other than marking to the element or element value, and the method is not particularly limited.

[0042] In S704, the information processing device 10 queries the rule tree information 107 to determine whether a node having a combination of the identified element and the acquired value exists in another persona. More specifically, the information processing device 10 reads out the alias set for the element of this combination, and identifies an element to which an alias having the same replacement value as the read out alias and an acceptable (i.e., similar) replacement value as the same replacement value according to a predetermined criterion is set. The identified element may exist in the rule tree of the same persona, or in the rule tree of another persona. Note that in the present invention, the information processing device 10 can also determine whether the same or similar combination exists by utilizing the combination of element and value itself without using an alias.

[0043] If it is determined that the same or similar combination exists as a result of the inquiry, in S705, the information processing device 10 marks one or more combinations of elements and element values ​​of the determined one or more personas as having acquired values. For example, in "C" in Fig. 8, a combination of a specific element and element value of the "body make" persona is determined to be the same combination, and is marked by the information processing device 10.

[0044] In S706, the information processing device 10 determines whether or not all elements and element values ​​existing between the top node and the target element of each persona marked in the processes of S703 and S705 (i.e., whose values ​​have been acquired) have been acquired. In the example of Fig. 8, all elements and values ​​existing between the top node and the target element of the "body make" persona have been acquired.

[0045] If it is determined that all have been acquired, in S707, the information processing device 10 determines that the value of the target element (i.e., the conclusion of the persona) has been acquired, and displays a comment (value value) of the target element value on the display of the user terminal 11. In "D" of Fig. 8, the information processing device 10 displays a comment, "From the perspective of body shaping, I would like you to eat a little more now," on the display of the user terminal 11.

[0046] In S708, the information processing device 10 queries the rule tree information 107 to determine whether elements that are the same as or similar to the acquired target element value exist in other personas. More specifically, the information processing device 10 reads out the alias set in the target element and determines whether there are elements to which an alias having the same replacement value as the read out alias and a replacement value that is acceptable as the same replacement value according to a predetermined criterion is set. If the determination results in the existence of elements, the information processing device 10 marks those elements as also having values ​​acquired. After this process, the process returns to S706.

[0047] If it is determined as a result of the inquiry process in S704 that no identical or similar combinations exist, and if it is determined as a result of the judgment process in S706 that all elements and element values ​​existing between the top node of each persona and the target element have not been acquired, in S709, the information processing device 10 selects the next topic.

[0048] In S709, the information processing device 10 compares the entropy values ​​of unmarked nodes among the elements of the rule tree of the persona in which the element identified in S704 exists, and selects an "element with high priority." The selected element becomes the next topic, and the process returns to S701, where the information processing device 10 executes a program associated with the element selected as the "element with high priority." After that, the process proceeds to S702.

[0049] Alternatively, in S709, the information processing device 10 compares the entropy values ​​of unmarked nodes among the elements of the rule trees of each persona determined in S706, and selects an "element with high priority." The selected element becomes the next topic, and the process returns to S701, where the information processing device 10 executes a program associated with the element selected as the "element with high priority." After that, the process proceeds to S702.

[0050] In S709, two processes were explained, but the difference is that the former compares the entropy values ​​within the same persona, while the latter compares the entropy values ​​within the same and other personas. In other words, in the latter case, there is a possibility that a topic will be selected from another persona.

[0051] In the above embodiment, the next topic element is determined based on the entropy value, but this is only one example. In the present invention, the information processing device 10 can select an element that is specified based on a predetermined requirement as the next topic element.

[0052] The predetermined requirements for identifying an element are roughly divided into two categories. One indicates that the information processing device 10 identifies an element based on one or more of character information, audio information, and image information that can be acquired via any sensor or device. The other indicates that the information processing device 10 identifies an element based on information obtained from a rule tree.

[0053] More specifically, regarding the former, the information processing device 10 can select the next topic element based on one or more of the following information acquired via any sensor or device: Compare the user's facial expression obtained from an arbitrary face recognition system with the facial expressions of various people learned in advance to determine whether the user's emotion is positive or negative, and decide whether to continue the dialogue using the rule tree for the current topic, or to have a dialogue that can transition to a rule tree for another topic (e.g., "By the way, regarding XX...", etc.). - Determine whether the user is likely to continue the conversation based on whether the character string entered by the user contains a predetermined word. If it is determined that the user is likely to continue the conversation, select an element of the currently selected rule tree that has not yet been given a value as the next topic. On the other hand, if it is determined that the user does not want to continue the conversation because, for example, the same conversation has continued for more than a predetermined number of times, a word has appeared that suggests the user is bored, or the length of the conversation sentence has become shorter than the previous input by more than a predetermined percentage, suggesting that interest is waning, change to a different rule tree.

[0054] More specifically, with respect to the latter, the information processing device 10 can select the next topical element based on information associated with the element other than the "entropy value." For each root from the top node of each of the multiple rule trees being discussed to the target element, calculate the ratio of the number of elements marked with an element value (the value is known) to the number of all elements, and select the rule tree with the root having the highest calculated ratio as the rule tree for the next topic. After that, provide the user with an arbitrary element from among the elements at the root of the selected rule tree as the next topic. In this way, the information processing device 10 can prioritize the selection of rule trees that are likely to lead to a conclusion.

[0055] The above-mentioned processing flow allows the information processing device 10 to use multiple personas to make recommendations for specific topics for users. For this reason, in the conventional technology, it was necessary to prepare a rule tree in advance in order to provide consulting to users on specific themes or problems, but according to the present invention, elements of different rule trees can be linked by using information associated with elements, such as an entropy value and an alias associated with an element, or information acquired via an arbitrary sensor or device, so that consulting for complex problems is also possible.

[0056] (Process to add persona) With reference to Figures 9 and 10, when it is found that a viewpoint is missing from an existing persona group, information processing for adding a new persona based on that viewpoint and modifying only related personas will be described. Figure 9 is a flow diagram explaining the processing for adding a new persona and modifying only related personas. Figure 10 is a diagram explaining the image of the processing for composing an existing persona group into a new persona group.

[0057] In this processing flow, we will take up a case where, after generating an existing group of personas, the food recommendations by the "food selection" persona do not include the perspective (element) of good or bad physical condition, and the persona needs to judge whether the physical condition is good or bad, so we will discuss a case where the group of personas needs to be reconstructed. For example, elements of the "physical condition" persona include, but are not limited to, chronic illness, stomach upset, calories, and nutrients.

[0058] In S901, the information processing device 10 acquires case data of the "Physical condition" persona to be added from the designer terminal 12 and stores it in the case data information 106. Taking into consideration the elements of the existing personas, the designer prepares case data of the "Physical condition" persona to be newly added and case data of the existing persona to be changed. As shown in Fig. 10, the elements included in the case data of the "Physical condition" persona are a mixture of common elements also included in the existing personas and newly added elements.

[0059] In S902, the information processing device 10 generates a rule tree for the “physical condition” persona based on the added case data, and stores it in the rule tree information 107.

[0060] In S903, the information processing device 10 acquires case data of the “meal selection” persona to be corrected from the designer terminal 12, and updates the data stored in the case data information 106.

[0061] In S904, the information processing device 10 regenerates the rule tree for the "meal selection" persona based on the updated case data.

[0062] As described above, in this processing flow, only personas related to the added element are newly generated or modified (replaced), and unrelated personas are not changed, so the scope of control is relatively small, and maintenance of each persona and confirmation of its contents become much easier. In other words, the new personas shown in Fig. 10 can be reconfigured with minimal addition and modification processing. Note that the order of the above flow may be changed, for example, the processing order of S902 and S903 may be changed.

[0063] (Process of converting output results into natural language) Fig. 11 explains a process flow for converting the results of a user dialogue using multiple personas into natural Japanese and outputting the results. If the elements of a rule tree generated from a case and the values ​​of a target element are output as is without editing, the language expression may not be natural. For this reason, the information processing device 10 can convert the output content presented to the user into a natural language expression according to the method shown in Fig. 11.

[0064] In the example described below, it is assumed that the user selects a general theme of the conversation, i.e., a persona that can be a target, when starting a conversation with a persona. For example, the user can select a general theme of the conversation, such as health, learning, shopping, or hobbies, and the information processing device 10 can select a target persona (rule tree).

[0065] In S1101, when a dialogue between a user and a persona is started, the information processing device 10 generates a real-time communication content table (referred to as an "RTC content table" in this specification) associated with the user dialogue, inputs a utterance at the start of the dialogue set in the top node of the selected persona (e.g., "Hello. What shall we talk about?") into the RTC content table, and provides it to the user terminal 11. As a result, the utterance at the start of the dialogue is displayed on the display of the user terminal 11.

[0066] In S1102, the user inputs a user input into the RTC content table via the user terminal 11 in response to a statement made at the start of the dialogue or in response to dialogue content received from the persona, and the information processing device 10 receives the user input into the RTC content table.

[0067] In S1103, the natural language recognition unit of the information processing device 10 acquires all elements and element values ​​of the persona group including the selected persona by a discovery acquisition request (GROUP), and analyzes the contents of the user input in the RTC content table based on the acquired information. The sentence of the user input may not match the information of the element, the target element, and their values ​​in the rule tree. Therefore, the natural language recognition unit segments the sentence of the user input according to a predetermined rule, determines whether it matches or is similar to the combination of the element and element value acquired for each segmented word, and identifies the element and element value related to each segmented word. In the present invention, the information processing device 10 can determine whether the combination of the element and element value acquired for each segmented word is similar based on the similar expression described in the similarity criterion table. The similar expression described in the similarity criterion table indicates an expression that can be considered similar even though the two expressions are not identical. The similar expression can indicate that the expression "A" is similar to the expressions "A1" and "A2", for example.

[0068] In S1104, the information processing device 10 determines whether the element identified in S1103 is a target element. An element type (target flag) can be set for the elements of the rule tree, and each element can be set to be a target element (conclusion) or an element (condition). If the result of the determination is that it is a target element, the process proceeds to S1106, and if it is not a target element, the process proceeds to S1105.

[0069] In S1105, the information processing device 10 performs natural language generation processing by the generation request function using the information of the identified element and element value, inputs a first natural language expression associated with the identified element and element value into the RTC content table, and provides it to the user terminal 11. The natural language generation processing by the generation request function of the information processing device 10 indicates that information of one element and element value is converted into a natural expression.

[0070] In S1106, the information processing device 10 causes the conversion unit of the information processing device 10 to execute a conversion process using the identified target element and target element value, and the contents input up to that point in the RTC content table. The conversion unit cooperates with the generative AI and uses functions provided in the generative AI to generate a second natural language expression based on the target element, target element value, and the input contents of the RTC content table. The "second natural language expression" refers to a natural language expression generated using functions provided in the generative AI.

[0071] The "first natural language expression" is obtained by converting the values ​​of elements and element values ​​into a natural language expression, and therefore can be processed by the generation request function of the information processing device 10. On the other hand, the "second natural language expression" is generated from the target elements and target element values ​​of one or more personas, as well as the input contents of the RTC content table, and therefore there is a large amount of information to be processed, and therefore the natural language expression is generated by using the functions provided in the generative AI.

[0072] When generating natural language expressions as described above, the information processing device 10 can generate language expressions by separating words and intentions. In detail, since the information processing device 10 cooperates with the generative AI and uses the functions of the generative AI, it can generate words that are in line with the times and in line with the local area in which the conclusions (target elements, values) of the personas are hot topics. In addition, the information processing device 10 can generate appropriate words by combining the conclusions (target elements, values) of multiple personas. For example, when the first persona concludes "encourage" and the second persona concludes "as if to admonish", the words generated are different from the words generated when the first persona concludes "encourage" and the second persona concludes "sympathize with the other person's position". Although appropriate words differ depending on each conclusion, the information processing device 10 can use the functions of the generative AI to convert the conclusions of multiple personas into appropriate expressions and output them to the RTC content table. Therefore, according to the present invention, it is not necessary to prepare different words for each combination of conclusions reached by multiple personas.

[0073] In S1107, the information processing device 10 inputs the conversion result, i.e., the second natural language expression, into the RTC content table and provides it to the user terminal 11. This enables the information processing device 10 to provide the user with a dialogue result in a more natural language expression for a user dialogue using multiple personas.

[0074] (Processing of adding the recognition result of the natural language recognition unit of the information processing device 10 to the case) The information processing device 10 can store the recognition result of the natural language recognition unit as a new case in the case data information 106. In the above, it has been described with reference to Figs. 4 and 5 that the information processing device 10 generates a persona (rule tree) from the case data. When generating a persona, all elements included in the case data are not used, and some elements with low information value are pruned. Fig. 12 is a diagram for explaining the concept of multiple personas generated based on the case data and elements not included in those personas. As shown in Fig. 12, the information processing device 10 generates a first persona, a second persona, etc. from the case data and stores them in the rule tree information 107, and stores information on the pruned elements in the auxiliary storage unit 103.

[0075] The discovery acquisition request described above is a function for acquiring a combination of elements and element values, and when the information processing device 10 uses a discovery acquisition request, it can control the range of information to be acquired by specifying the following four parameters. Note that in S1103, since the discovery acquisition request (GROUP) is used, all elements and element values ​​of the persona group including the persona selected in the dialogue are acquired.

[0076] [Table 1]

[0077] 13 explains the concept of the process in which the information processing device 10 adds a recognition result as case data based on the original data of a new case and combinations of all elements and element values ​​acquired by a discovery acquisition request (ALL). The information processing device 10 determines which element and element value acquired corresponds to the input data from the original data of the case, and stores the combination of elements and cases as the determination result (recognition result) in the case data information 106. Through this process, case data is accumulated and the personas are brushed up.

[0078] Although the principles of the present invention have been described above with reference to exemplary embodiments, various embodiments that change the configuration and details can be realized without departing from the gist of the present invention. That is, the present invention can be embodied as, for example, a system, an apparatus, a method, a program, or a storage medium. [Explanation of symbols]

[0079] 10. Information processing device 11 User terminal 12 Designer Terminal 101 Control section 102 Main memory 103 Auxiliary storage 104 IF Section 105 Output section 106 Case Data Information 107 Rule Tree Information 111 Elements 112 Element Value

Claims

1. A processor; A storage unit for storing a plurality of rule trees; An information processing device comprising: The processor, Marking a first element being discussed in a first rule tree as having a value acquired; Querying the storage to determine whether there is a second element having the same and similar alias as the first element that is marked; If the second element is present, marking the second element as having a value obtained; For each of a first rule tree associated with the first marked element and a second rule tree associated with the second marked element, determining whether all elements present between a top node and a target element of the tree traversed as a result of user interaction are marked; selecting, on a condition that all of the elements are unmarked, an element identified based on a predetermined criterion from among the unmarked elements of the first rule tree and the unmarked elements of the second rule tree as a next topic; An information processing device configured as above.

2. The processor, Run the program associated with the element selected as the next topic; providing a user with a value associated with the selected element; obtaining a value from the user and marking the selected element as having a value obtained; The information processing apparatus of claim 1 further configured to:

3. The processor, providing the user with a value to be associated with the target element, provided that all elements between the top node and the target element in the tree traversed as a result of the user interaction are marked; querying the storage to determine whether a third element exists that has a same or similar alias as the target element, and if so, marking the third element as having a value obtained; The information processing apparatus of claim 1 further configured to:

4. The processor, selecting, on a condition that the second element is not present, an element identified based on a predetermined criterion from among the unmarked elements of the first rule tree as a next topic; The information processing apparatus of claim 1 further configured to:

5. A method executed by an information processing device, comprising: The information processing device includes a storage unit that stores a plurality of rule trees, and the method includes: marking a first element being discussed on a first rule tree as having a value acquired; Querying the storage to determine whether there is a second element that has the same and similar alias as the first element that is marked; If the second element is present, marking the second element as having been valued; determining whether all elements present between a top node and a target element of the tree traversed as a result of user interaction are marked, for each of a first rule tree associated with the first marked element and a second rule tree associated with the second marked element; selecting, on a condition that all of the elements are unmarked, an element identified based on a predetermined criterion from among the unmarked elements of the first rule tree and the unmarked elements of the second rule tree as a next topic; A method for providing the above.

6. Executing a program associated with the element selected as the next topic; providing a user with a value associated with the selected element; obtaining a value from the user and marking the selected element as having a value obtained; The method of claim 5 , further comprising: providing a user with a value associated with the target element, provided that all elements between the top node and the target element in a tree traversed as a result of user interaction are marked; querying the storage to determine whether a third element exists that has a same and similar alias as the target element, and if so, marking the third element as having a value obtained; The method of claim 5 , further comprising:

8. 6. The method of claim 5, further comprising: selecting, on the condition that the second element is not present, an element identified based on a predetermined criterion among unmarked elements of the first rule tree as a next topic.

9. A program causing a computer to execute the method according to any one of claims 5 to 8.

Images