Question answering device, question answering method, and program

Abstract

In a question answering device (70), an acquisition means (71) acquires a query. A start point phrase determination means (72) determines a start point phrase corresponding to the query from a plurality of phrases included in causal relationship data. An inference type determination means (73) determines an inference type on the basis of the query. A search means (74) searches for a phrase by following a causal relationship included in the causal relationship data from the start point phrase according to a search procedure corresponding to the determined inference type. An output means (75) outputs, as an answer to the query, a phrase obtained by the search.

Description

Technical Field

[0001] This disclosure relates to question - answering technology.

Background Art

[0002] There is known a question - answering system that receives a natural - language question from a user and returns an answer thereto. Patent Document 1 discloses a question - answering device that outputs an answer to a question sentence in consideration of the causal relationship between the question sentence and answer candidates.

Prior Art Documents

Non - Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] When generating an answer to a question using causal - relationship knowledge, it is required to efficiently generate and output an answer to the question based on a huge amount of causal - relationship knowledge.

[0005] One object of this disclosure is to provide a question - answering device capable of efficiently generating and outputting an answer to a question based on causal - relationship knowledge.

Means for Solving the Problems

[0006] In one aspect of this disclosure, the question - answering device query acquisition means for acquiring a query, starting - phrase determination means for determining a starting phrase corresponding to the query from a plurality of phrases included in causal - relationship data, inference - type determination means for determining an inference type based on the query, A search means that searches for phrases by tracing the causal relationships contained in the causal relationship data from the starting phrase in a search procedure corresponding to the determined inference type, An output means that outputs the phrase obtained by the search as the answer to the query, It is equipped with.

[0007] In other respects of this disclosure, a computer-based question answering method is: Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, The phrase obtained through the search is output as the answer to the query.

[0008] In yet another aspect of the present invention, program teeth, Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, The computer is instructed to execute a process that outputs the phrase obtained through the aforementioned search as the answer to the query. 。 [Brief explanation of the drawing]

[0009] [Figure 1] The overall configuration of the question answering system according to the first embodiment is shown. [Figure 2] This is a block diagram showing the hardware configuration of a question answering device. [Figure 3] This is a block diagram showing the functional configuration of a question answering device. [Figure 4] Schematically shows an example of causal relationship knowledge stored in the causal relationship database. [Figure 5] Shows an example of causal inference. [Figure 6] Shows an example of inverse causal inference. [Figure 7] Shows an example of problem - solving inference. [Figure 8] Shows an example of solution - problem inference. [Figure 9] It is a flowchart of the inference process by the question - answering device of the first embodiment. [Figure 10] It is a block diagram showing the functional configuration of the question - answering device of the second embodiment. [Figure 11] It is a flowchart of the inference process by the question - answering device of the second embodiment. [Figure 12] Shows an example of the display of the answer graph output by the question - answering device of the second embodiment. [Figure 13] Shows another example of the display of the answer graph output by the question - answering device of the second embodiment. [Figure 14] It is a block diagram showing the configuration of the medical knowledge generation system according to the third embodiment. [Figure 15] It is a block diagram showing the configuration of the question - answering system according to the third embodiment. [Figure 16] It is a block diagram showing the functional configuration of the question - answering device of the fourth embodiment. [Figure 17] It is a flowchart of the process by the question - answering device of the fourth embodiment.

MODE FOR CARRYING OUT THE INVENTION

[0010] <0​​​​​Figure 1 shows the overall configuration of the question answering system according to the first embodiment. The question answering system 1 comprises a user terminal device 5 and a question answering device 100. The user operates the terminal device 5 to send a question (hereinafter referred to as "query") to the question answering device 100. The query is written in natural language. The question answering device 100 receives the query from the user. The question answering device 100 generates an answer to the query using causal relationship knowledge that defines the relationships between multiple causally related phrases and sends it to the user's terminal device 5.

[0011] (Basic principle) Next, the basic principle of the question answering device 100 will be explained. The question answering device 100 searches for answers to user queries based on pre-prepared causal relationship knowledge. Causal relationship knowledge defines the causal relationships between multiple sentences (hereinafter also referred to as "phrases") written in natural language. First, the question answering device 100 determines a starting phrase from among the multiple phrases included in the causal relationship knowledge that corresponds to the user's query. Then, the question answering device 100 searches for phrases that have a causal relationship with the user's query by tracing the causal relationships included in the causal relationship knowledge from the starting phrase, and generates an answer to the query. In addition to causal relationships, the question answering device 100 may also search for phrases by tracing specific semantic relationships, which will be described later. In this way, the question answering device 100 can generate answers that have a causal relationship with the user's query based on pre-prepared causal relationship knowledge. The process of deriving an answer to a user's query based on causal relationship knowledge is called "inference".

[0012] (Hardware configuration) Figure 2 is a block diagram showing the hardware configuration of the question answering device 100. As shown in the figure, the question answering device 100 includes an interface (IF) 12, a processor 13, a memory 14, a recording medium 15, and a database (DB) 16.

[0013] IF12 obtains user-entered queries by communicating with an external device. IF12 also transmits the answers generated by the question answering device 100 to the external device. Specifically, when a user sends a query to the question answering device 100 using terminal device 5, the question answering device 100 receives the query through IF12 and transmits the corresponding answer to the query to the user's terminal device 5 via IF12.

[0014] The processor 13 is a computer such as a CPU (Central Processing Unit) and controls the entire question answering device 100 by executing a pre-prepared program. Specifically, the processor 13 can be a CPU, GPU (Graphics Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof. Specifically, the processor 13 performs inference processing to generate answers to queries.

[0015] Memory 14 refers to ROM (Read Only Memory) and RAM (Random Memory). It consists of components such as Access Memory. Memory 14 stores various programs executed by the processor 13. Memory 14 is also used as working memory while the processor 13 is executing various processes.

[0016] The recording medium 15 is a non-volatile, non-temporary recording medium such as a disk-shaped recording medium or semiconductor memory, and is configured to be detachable from the question answering device 100. The recording medium 15 stores various programs that the processor 13 executes. When the question answering device 100 performs various processes, the programs stored on the recording medium 15 are loaded into the memory 14 and executed by the processor 13.

[0017] DB16 stores causal knowledge used to generate answers to queries. DB16 may also store machine learning models and other information used in the inference process described later.

[0018] In addition to the above, the question answering device 100 may also be equipped with a display device such as a liquid crystal display or a projector, and an input device such as a keyboard or mouse. These display devices and input devices are used, for example, by the administrator of the question answering device 100 to perform necessary management.

[0019] (Functional Configuration) Figure 3 is a block diagram showing the functional configuration of the question answering device 100. Functionally, the question answering device 100 comprises a causal relationship DB 21, a semantic relationship determination unit 22, a starting phrase determination unit 23, an inference type determination unit 24, and an inference unit 25.

[0020] The causal relationship DB21 stores causal relationship knowledge between multiple phrases. Causal relationship knowledge is an example of causal relationship data, and is knowledge data extracted from a large amount of literature, web documents, etc. In this embodiment, causal relationship knowledge is defined as consisting of the following elements that have a causal relationship. • "Natural language sentences (cause phrases) that express cause" • "Natural language sentences (result phrases) that express the result" Causal relationship knowledge is prepared in advance as multiple pairs of cause phrases and effect phrases (hereinafter also referred to as "causal relationship pairs") and stored in the causal relationship DB21.

[0021] In this embodiment, "causal relationship" refers to a correspondence such as "if one occurs, then the other occurs." Therefore, in this embodiment, the causal relationship includes the following relationships.

[0022] (1) Causal relationship in the narrow sense In the narrow sense, a causal relationship refers to a relationship where one cause leads to another. For example, "The birth rate declines." → "The population ages." (2) Context Context refers to a temporal relationship where one event occurs before the other. For example, "Go to a restaurant." → "Order a meal." (3) Implication An implication relationship is a relationship in which one statement implies the other. For example, "Manage the condition of the fields with AI." → "Turn agriculture into IT."

[0023] As described above, the causal relationship DB21 stores multiple causal relationship pairs, but additional information such as the following may also be stored in the causal relationship DB21. (i) Source information for causal relationships Source information for causal relationships indicates the source of information, such as literature or web documents, in which a particular causal relationship pair is described. Source information for causal relationships can be used to show the basis for the selection of an answer when presenting an answer to a query to the user. (ii) Subdivision of causal relationships A detailed classification of causal relationships is a classification that shows a more detailed relationship between the phrases that make up a causal relationship pair. Examples include "the relationship between a problem and its solution (problem-solving relationship)" and "temporal relationships." In addition, the relationships included in "causal relationships" in this embodiment (causal relationships in the narrow sense, temporal relationships, and implicational relationships) as mentioned above may also be used as detailed classifications.

[0024] Figure 4 schematically shows an example of causal knowledge stored in the causal relationship DB21. For ease of explanation, Figure 4 shows the causal knowledge in a graph structure. Each phrase F1 to F6 corresponds to a node in the graph structure, and the causal relationship between phrases is indicated by solid arrows (edges). For example, phrases F1 and F2 form a causal relationship pair: "Aging is progressing in rural areas." → "Drones are being introduced to rural areas." Note that Figure 4 is an excerpt of a portion of the causal knowledge containing numerous causal relationship pairs.

[0025] Furthermore, the data structure of the causal relationship knowledge stored in the causal relationship DB21 is not limited to the graph structure described above. For example, the causal relationship DB21 may store causal relationship knowledge as a list of numerous causal relationship pairs.

[0026] The semantic relationship determination unit 22 extracts pairs of phrases that correspond to a specific semantic relationship (hereinafter also referred to as "specific semantic relationship pairs") from among multiple phrases included in the causal relationship knowledge stored in the causal relationship DB 21. Here, "specific semantic relationship" refers to a relationship in which the meanings of the phrases are similar, and specifically includes implication relationships, synonym relationships, and similarity relationships. An implication relationship is a relationship in which one of the two phrases implies the other. A synonym relationship is a relationship in which the meanings of the two phrases are the same, and a similarity relationship is a relationship in which the meanings of the two phrases are similar.

[0027] The reason for extracting specific semantic relationship pairs from multiple phrases contained in causal knowledge is as follows: As mentioned earlier, the causal relationship pairs that constitute causal knowledge are phrase pairs written in natural language, and therefore causal knowledge contains multiple phrases that have implication relationships or multiple phrases that describe substantially the same or similar content. Therefore, by flexibly linking such semantically similar phrases to expand the causal knowledge, it becomes possible to infer answers from queries using a broader range of causal knowledge.

[0028] In Figure 4, the dashed arrows indicate edges connecting specific semantic relationship pairs. In the example in Figure 4, phrase F2, "Drones are introduced to rural areas," and phrase F5, "Pesticides are sprayed using drones," have an implication relationship and constitute a specific semantic relationship pair. Thus, although there is no causal relationship between phrases F2 and F5 in the original causal knowledge, phrases F2 and F5 are defined as a specific semantic relationship pair. Then, when the question answering device 100 performs inference based on a query, it searches for an answer to the query by following causal relationships and specific semantic relationships. In this way, in the example in Figure 4, the question answering device 100 can search by following more phrases, such as phrase F1 → F2 → F5 → F6. This makes it possible to infer an answer to a query using a broader range of causal knowledge.

[0029] For example, the semantic relationship determination unit 22 can determine whether two phrases have a specific semantic relationship using the feature vectors of the phrases to be determined. For instance, the semantic relationship determination unit 22 converts each of the two phrases into feature vectors and calculates the similarity or distance between them. Then, if the similarity or distance between the two feature vectors is below a predetermined threshold, the semantic relationship determination unit 22 determines that the phrases have a specific semantic relationship. The semantic relationship determination unit 22 outputs the extracted specific semantic relationship pair to the inference unit 25.

[0030] The starting phrase determination unit 23 refers to the causal relationship knowledge stored in the causal relationship DB 21 and determines the starting phrase corresponding to the query entered by the user. The "starting phrase" is the phrase that serves as the starting point when tracing causal relationship knowledge based on the query entered by the user in the inference of the answer to the query. Specifically, the starting phrase determination unit 23 analyzes the meaning of the user's query. Then, the starting phrase determination unit 23 determines the starting phrase to be the phrase that is closest in meaning to the user's query from among several phrases included in the causal relationship knowledge. For example, if the user's query was "What happens when rural areas age?", the starting phrase determination unit 23 searches the causal relationship knowledge for a phrase that is substantially the same in meaning as this query, or a phrase that is closer in meaning than a predetermined criterion. Then, if the starting phrase determination unit 23 finds the phrase "aging rural areas" in the causal relationship knowledge stored in the causal relationship DB 21, for example, it determines that the phrase to be the starting phrase. In one example, the starting phrase determination unit 23 may determine a starting phrase as a phrase in which the similarity of feature vectors between phrases is higher than a predetermined threshold value, or a phrase in which the distance between feature vectors is closer than a predetermined threshold value, similar to the semantic relationship determination unit 22 described above. The starting phrase determination unit 23 outputs the determined starting phrase to the inference unit 25.

[0031] Furthermore, the starting phrase determined by the starting phrase determination unit 23 is not limited to one starting phrase per query; there may be multiple starting phrases. For example, if the user's query is "rural issues," the starting phrase determination unit 23 may select several phrases from among the multiple phrases included in the causal relationship knowledge that are closest to "rural issues" as the starting phrase.

[0032] The inference type determination unit 24 determines the inference type based on the query entered by the user. Here, "inference type" refers to the procedure for searching for phrases by tracing causal knowledge when inferring an answer corresponding to a query based on causal knowledge. Specifically, the following are examples of inference types.

[0033] (A) Causal inference Causal inference is a type of inference that responds to queries such as "What happens if XX occurs?", and it is a method of searching for causal relationships in a forward direction within causal knowledge, tracing causal relationships and specific semantic relationships. Forward direction refers to, for example, the direction from "cause" to "effect".

[0034] Figure 5 shows an example of causal inference. In the figure, solid arrows indicate causal relationships, and dashed arrows indicate problem-solving relationships. Causal inference infers events that occur as a result of a query. In the example in Figure 5, based on the query "Aging is progressing in rural areas," the phrase "Food self-sufficiency rates are declining" is obtained. Furthermore, based on the phrase "Food self-sufficiency rates are declining," the phrase "Household food expenses are rising" is obtained.

[0035] (B) Reverse causal inference Reverse causal inference is a type of inference that responds to queries such as "What caused XX?", and it is a method of exploration that traces causal relationships and specific semantic relationships in reverse. The reverse direction refers to, for example, the direction from "effect" to "cause".

[0036] Figure 6 shows an example of reverse causal inference. In the figure, solid arrows indicate causal relationships, and dashed arrows indicate problem-solving relationships. Reverse causal inference infers the events that cause a query. In the example in Figure 6, based on the query "Aging is progressing in rural areas," the phrase "Birth rates are declining" is obtained. Furthermore, based on the phrase "Birth rates are declining," the phrase "Marriage is getting later" is obtained.

[0037] (C) Combinatorial inference Combinatorial inference is a method that combines the above-mentioned causal inference and inverse causal inference, and includes, for example, the following methods.

[0038] (C1) Problem-solution reasoning Problem-solution reasoning is a type of reasoning that responds to queries such as "How can we solve XX?". It is a search method that traces causal relationships in the forward direction from the problem to the solution, and then traces causal and specific semantic relationships in the reverse direction. This allows us to obtain a solution to a problem, and then obtain other problems that are simultaneously solved by that solution, as well as situations and phenomena that arise as a result of that solution.

[0039] Figure 7 shows an example of problem-solution reasoning. In the figure, solid arrows indicate causal relationships, and dashed arrows indicate problem-solving relationships. Problem-solution reasoning infers solutions from a query, and then infers events that trigger the realization of those solutions. In the example in Figure 7, the query "Aging is progressing in rural areas." leads to the phrase "Agricultural automation is progressing." Furthermore, based on the phrase "Agricultural automation is progressing," the phrase "Drones will be introduced into agriculture" is obtained, and based on the phrase "Drones will be introduced into agriculture," the phrase "Legal frameworks regarding drones are progressing" is obtained.

[0040] (C2) Solving-problem reasoning Solution-problem reasoning is a type of reasoning that responds to queries such as "What problem will be solved if XX is done?". It is a search method that traces causal relationships backward from the solution to the problem, and then traces causal and specific semantic relationships forward from there. This allows, for example, to obtain a problem that is solved by a certain solution, and then to obtain another solution that solves that problem.

[0041] Figure 8 shows an example of solution-problem reasoning. In the figure, solid arrows indicate causal relationships, and dashed arrows indicate problem-solving relationships. Solution-problem reasoning infers the problem to be solved by the query, and then infers the events that will be resolved as a result of solving that problem. In the example in Figure 8, the phrase "Aging is progressing in rural areas" is obtained based on the query "Aging is progressing in rural areas." Furthermore, based on the phrase "Aging is progressing in rural areas," the phrase "The agricultural workforce is declining" is obtained.

[0042] (D) Classification designation inference Subcategory-specific inference is a search method that limits the causal relationships used in inference to those of a specific subcategory. In other words, subcategory-specific inference searches by considering or excluding causal relationships belonging to a specific subcategory from among multiple causal relationships included in the causal relationship knowledge. For example, as mentioned above, suppose the causal relationship knowledge stored in the causal relationship DB21 includes three subcategories: "causal relationships in the narrow sense," "problem-solving relationships," and "temporal relationships." If the user specifies "consider only problem-solving relationships" as a condition for query-based inference, the question answering device 100 searches by following only problem-solving relationships and their related specific semantic relationships from the causal relationship knowledge stored in the causal relationship DB21. Conversely, if the user specifies "exclude problem-solving relationships" as a condition for query-based inference, the question answering device 100 searches by following only causal relationships other than problem-solving relationships and their related specific semantic relationships from the causal relationship knowledge stored in the causal relationship DB21.

[0043] Next, the method by which the inference type determination unit 24 determines the inference type will be described. The inference type determination unit 24 determines the inference type by text analysis of the query entered by the user. Specifically, in the first method, the inference type determination unit 24 determines the inference type using a rule-based machine learning model. For example, a machine learning model is generated that determines the inference type based on words and symbols contained in the phrase that makes up the query, and this is used to determine the inference type. In the second method, the inference type determination unit 24 generates a classification model that takes the query phrase as input and outputs the corresponding inference type, and this is used to determine the inference type. This classification model may, for example, use a neural network. In the third method, the inference type determination unit 24 determines the inference type based on whether or not a specific word (for example, "why," "how," etc.) or symbol ("?", etc.) is included at the beginning or end of the sentence that makes up the query. The inference type determination unit 24 outputs the inference type determined by any of the methods to the inference unit 25.

[0044] The inference unit 25 infers an answer to the query using the starting phrase determined by the starting phrase determination unit 23 and the inference type determined by the inference type determination unit 24. Specifically, the inference unit 25 searches for phrases that have a causal relationship with the query by tracing causal relationships and specific semantic relationships starting from the starting phrase in causal relationship knowledge including specific semantic relationship pairs, as illustrated in Figure 4. The inference unit 25 then outputs the phrases of nodes where causal relationships and specific semantic relationships can no longer be traced (hereinafter also referred to as "terminal nodes") as the answer to the query. As mentioned earlier, if the starting phrase determination unit 23 determines multiple starting phrases for the query, the inference unit 25 can search for each phrase individually starting from each starting phrase and output the phrases of the terminal nodes corresponding to each search as the answer.

[0045] (Inference processing) Figure 9 is a flowchart of the inference process performed by the question answering device 100 in the first embodiment. This process is achieved by the processor 13 shown in Figure 2 executing a pre-prepared program and operating as each element shown in Figure 3.

[0046] First, the question answering device 100 obtains the query entered by the user (step S21). Next, the starting phrase determination unit 23 determines the starting phrase corresponding to the entered query from the causal relationship knowledge stored in the causal relationship DB 21 (step S22). Next, the inference type determination unit 24 determines the inference type corresponding to the entered query (step S23).

[0047] Next, the inference unit 25 searches for phrases by tracing the causal relationships and specific semantic relationships included in the causal relationship knowledge, according to the inference type determined in step S23, starting from the starting phrase determined in step S22 (step S24). Specifically, the inference unit 25 traces causal relationships based on the causal relationship knowledge. At the same time, the semantic relationship determination unit 22 determines whether or not there is a phrase that has a specific semantic relationship with the phrase at the current position in the search. If there is a phrase that has a specific semantic relationship with the phrase at the current position, the inference unit 25 also traces that specific semantic relationship in its search. That is, the inference unit 25 searches for phrases while also considering the specific semantic relationships extracted by the semantic relationship determination unit 22. Finally, when the inference unit 25 has run out of causal relationships and specific semantic relationships to trace, that is, when the current position of the search has reached a terminal node, it outputs the phrase of that terminal node as the answer to the query (step S25). The inference process then ends.

[0048] As described above, in the first embodiment, in addition to pre-prepared causal relationship knowledge, the system infers answers to queries by targeting phrases with specific semantic relationships such as implication, synonymy, and similarity, making it possible to infer answers based on a larger amount of causal relationship knowledge. Furthermore, the system determines the starting phrase and inference type based on the query entered by the user, and searches for phrases from the determined starting phrase according to the determined inference type, making it possible to infer answers in a way that is suitable for the entered query.

[0049] <Second Embodiment> Next, a second embodiment of the present disclosure will be described. In the first embodiment, the question answering device 100 performs inference for a query while extracting specific semantic relationship pairs using the semantic relationship determination unit 22. Instead, in the second embodiment, the question answering device extracts specific semantic relationships in advance from the causal relationship knowledge stored in the causal relationship DB and creates a relationship graph that includes causal relationships and specific semantic relationships. In addition, in the second embodiment, the question answering device outputs a response graph, which will be described later, as an answer to a query. Aside from these points, the second embodiment is basically the same as the first embodiment. Specifically, the overall configuration and hardware configuration of the question answering device 100x of the second embodiment are the same as the overall configuration of the first embodiment shown in Figure 1 and the hardware configuration shown in Figure 2, so a description will be omitted.

[0050] (Functional Configuration) Figure 10 is a block diagram showing the functional configuration of the question answering device 100x according to the second embodiment. As shown in the figure, the question answering device 100x of the second embodiment includes a causal relationship DB 31, a semantic relationship determination unit 32, a starting phrase determination unit 33, an inference type determination unit 34, an inference unit B 35, and a relationship graph 36. Note that the causal relationship DB 31, the starting phrase determination unit 33, and the inference type determination unit 34 are basically the same as the causal relationship DB 21, the starting phrase determination unit 23, and the inference type determination unit 24 in the first embodiment, so their explanation is omitted.

[0051] As a preprocessing step, the semantic relationship determination unit 32 extracts specific semantic relationship pairs from the entire causal relationship knowledge stored in the causal relationship DB 31. Then, the semantic relationship determination unit 32 generates a relationship graph 36 by adding the extracted specific semantic relationship pairs to the causal relationship knowledge. The generated relationship graph 36 has a graph structure that includes causal relationships and specific semantic relationships, as illustrated in Figure 4. That is, the relationship graph 36 is a directed graph in which each phrase is represented by a node and causal relationships or specific semantic relationships are represented by edges. In this way, in the second embodiment, a relationship graph 36 is generated by adding specific semantic relationship pairs to the causal relationship knowledge as a preprocessing step, and when a user inputs a query, the system refers to this relationship graph 36 to infer the answer. Therefore, the computational load of inference for query input can be reduced, and the time required to output the answer can be shortened.

[0052] When a user enters a query, the starting phrase determination unit 33 determines the starting phrase, and the inference type determination unit 34 determines the inference type, similar to the first embodiment. Then, the inference unit 35 searches for a phrase in the relationship graph 36, starting from the starting phrase and tracing causal relationships and specific semantic relationships. In the first embodiment, the inference unit 25 outputs the phrase of the terminal node where causal relationships and specific semantic relationships can no longer be traced as the answer to the query. In contrast, in the second embodiment, the inference unit 35 outputs a graph (referred to as the "answer graph") showing the portion of the phrase search where causal relationships and specific semantic relationships were traced as the answer to the query.

[0053] (Inference processing) Figure 11 is a flowchart of the inference process by the question answering device 100x in the second embodiment. This process is realized by the processor 13 shown in Figure 2 executing a pre-prepared program and operating as each element shown in Figure 10. In the second embodiment, as a preprocessing step, the semantic relationship determination unit 32 extracts specific semantic relationship pairs between phrases included in the causal relationship DB 31 and generates a relationship graph 36 that includes causal relationships and specific semantic relationships.

[0054] First, the question answering device 100x obtains the query entered by the user (step S31). Next, the starting phrase determination unit 33 determines the starting phrase corresponding to the entered query from the causal relationship knowledge stored in the causal relationship DB 31 (step S32). Next, the inference type determination unit 34 determines the inference type corresponding to the entered query (step S33).

[0055] Next, the inference unit 35 searches for phrases by tracing causal relationships and specific semantic relationships in the relationship graph 36, starting from the starting phrase determined in step S32, according to the inference type determined in step S33 (step S34). Then, if there are no more causal relationships or specific semantic relationships to trace, the inference unit 35 outputs the portion it has traced so far from the relationship graph 36 as the answer graph (step S25). The answer graph is then presented to the user, and the inference process ends.

[0056] (Example display) Figure 12 shows an example of the display of the answer graph output by the inference unit 35. In this example, the user inputs the query, "What happens when rural areas age?", and the answer graph 40 for this query is displayed. The answer graph 40 is an excerpt of the relationship graph 36 used by the inference unit 35 for inference, showing the portion that the inference unit 35 followed during phrase search. Specifically, in the example in Figure 12, for the query, "What happens when rural areas age?", the starting phrase determination unit 33 determines node 41a as the starting phrase, and the inference unit 35 proceeds from node 41a through edge 42a indicating a causal relationship to node 41b, then through edge 42b indicating a specific semantic relationship to node 41c, and further through edge 42c indicating a causal relationship to terminal node 41d, thus ending the search. In this way, in the second embodiment, the answer graph showing the portion of the relationship graph 36 that the inference unit 35 followed during inference is displayed as the answer to the query, so the user can understand how the answer was generated.

[0057] Figure 13 shows another example of the display of the response graph output by the inference unit 25. In this example, the user enters the query "What will happen with agricultural DX?", and the response graph 50 for this query is displayed. Specifically, in the example in Figure 13, the inference unit 35 first determines the starting phrase 51x based on the query "What will happen with agricultural DX?". Next, the inference unit 35 searches for nodes 51a and 51b that have a causal relationship with the starting phrase 51x and displays those phrases. Furthermore, the inference unit 35 searches for nodes 51c to 51f that have a problem-solving relationship with the starting phrase 51x, and further searches for nodes 51g and 51h that have a causal relationship with node 51e, and displays those phrases.

[0058] As mentioned above, if the causal relationship knowledge stored in the causal relationship DB31 includes source information for each causal relationship as additional information, the question answering device 100x may display the corresponding source information according to the user's specification. For example, if the user clicks on an edge indicating a causal relationship in the displayed answer graph, the question answering device 100x may display the source information of the causal relationship knowledge corresponding to that edge (for example, the URL of the web document describing that information) in a pop-up or similar. In the display example in Figure 13, as a result of the user clicking on edge E51d connecting nodes 51x and 51d, the source (information source) of the causal relationship knowledge indicated by edge E51d is displayed in answer 52.

[0059] Furthermore, if the causal relationship knowledge stored in the causal relationship DB31 includes subcategories of each causal relationship as additional information, the question answering device 100x may display these subcategories according to the user's specifications. For example, if the user clicks on an edge indicating a causal relationship in the displayed answer graph, the question answering device 100x may display the subcategory of the causal relationship corresponding to that edge (e.g., "temporal relationship") in a pop-up window or similar. In the example display in Figure 13, the user clicks on edge E51d connecting nodes 51x and 51d, and as a result, the answer 52 displays the subcategory of the causal relationship of edge E51d, "problem-solving relationship".

[0060] Furthermore, the user may be allowed to select a node included in the response graph and issue a search instruction for a phrase corresponding to the query. In this case, the question answering device 100x performs the search using the node selected by the user as the starting phrase.

[0061] Furthermore, the system may allow the user to input an evaluation of the displayed response graph or of the question answering device 100x. For example, if the inference type determination unit 34 determines the inference type based on rules, the rules used to determine the inference type may be modified based on the user's evaluation input. Also, if the inference type is determined using a machine learning model, the user's evaluation input may be used as training data, and the model used to determine the inference type may be retrained.

[0062] Furthermore, users may be allowed to input modifications to the causal relationship knowledge in the displayed response graph. For example, users may be allowed to add new edges or modify the direction of edges (arrows) on the displayed response graph. This makes it possible to modify the causal relationship knowledge used for inference based on user input. Also, when generating a causal relationship graph that shows causal relationship knowledge using machine learning, the user input described above may be used as additional training data to retrain the causal relationship graph.

[0063] As described above, in the second embodiment, a relational graph 36 is prepared in advance, which includes phrases with specific semantic relationships such as implication, synonymy, and similarity, in addition to pre-prepared causal relationship knowledge. This reduces the processing load from when the user inputs a query until the answer is output, making it possible to perform inference in a short time. Furthermore, since the answer graph showing the causal relationships and specific semantic relationships used in the inference is displayed as the answer to the query, the user can learn about the process and basis for obtaining the answer.

[0064] <Third Embodiment> Next, a third embodiment of the present disclosure will be described. The third embodiment applies the question answering device of the first and second embodiments to medical question answering.

[0065] Figure 14 shows the configuration of the medical knowledge generation system according to the third embodiment. As shown in the figure, the medical knowledge generation system 200x comprises a medical knowledge generation device 201 and a medical knowledge database 202. The medical knowledge generation device 201 receives the voice of a doctor or nurse during the diagnosis of a patient. The medical knowledge generation device 201 is equipped with a medical knowledge generation AI. The medical knowledge generation AI has the function of converting the input voice into text and summarizing its contents, the function of automatically generating a medical record based on the summarized text, and the function of constructing causal relationship knowledge from the medical record. Using the medical knowledge generation AI, the medical knowledge generation device 201 generates causal relationship knowledge related to medicine from the voice input by doctors and nurses and stores it as medical knowledge in the medical knowledge database 202.

[0066] Specifically, the medical knowledge generation device 201 generates medical knowledge using a medical knowledge generation AI as follows: When a doctor makes a diagnosis to a patient, the medical knowledge generation AI records the doctor's or nurse's voice. The medical knowledge generation AI converts the doctor's or nurse's voice into text. This text reflects the content of the diagnosis made by the doctor to the patient. Next, the medical knowledge generation AI summarizes the text obtained through the conversion. This summary is a concise summary of the diagnosis made by the doctor to the patient. Next, the medical knowledge generation AI automatically generates a patient's medical record based on the text obtained through the summary. This medical record contains basic information about the patient, the content of the diagnosis, treatment methods, etc.

[0067] Next, the medical knowledge generation AI constructs causal relationship knowledge from the generated medical records using a predetermined machine learning algorithm. In this process, the causal relationship knowledge may include not only past user medical records but also publicly available medical information. In this way, causal relationship knowledge related to medical care is generated and stored as medical knowledge in the medical knowledge DB202.

[0068] Figure 15 shows the configuration of the question-answering system according to the third embodiment. The question-answering system 200y supports the diagnosis of patients performed by physicians by performing AI diagnosis using a question-answering device. The question-answering system 200y is an application of the question-answering device 100 according to the first embodiment or the question-answering device 100x according to the second embodiment.

[0069] In diagnosing a patient, the physician inputs a query into the question answering device 100 or 100x using the terminal device 5. The question answering device 100 or 100x uses the medical knowledge DB 202, which stores causal knowledge related to medical care, to generate an answer to the physician's query and transmit it to the physician's terminal device 5. Using the technology of the first or second embodiment, the question answering device 100 or 100x infers and optimizes the cause of the patient's symptoms and the diagnosis result using past user medical records and publicly available medical information, and transmits this as an answer to the physician's terminal device 5. The physician can make a diagnosis of the patient by referring to the answer from the question answering device 100 or 100x.

[0070] <Fourth Embodiment> Figure 16 is a block diagram showing the functional configuration of the question answering device according to the fourth embodiment. The question answering device 70 comprises an acquisition means 71, a starting phrase determination means 72, an inference type determination means 73, a search means 74, and an output means 75.

[0071] Figure 17 is a flowchart of the processing performed by the question answering device 70 of the fourth embodiment. The acquisition means 71 acquires the query (step S71). The starting phrase determination means 72 determines a starting phrase corresponding to the query from a plurality of phrases included in the causal relationship data (step S72). The inference type determination means 73 determines the inference type based on the query (step S73). The search means 74 searches for phrases by tracing the causal relationships included in the causal relationship data from the starting phrase using a search procedure corresponding to the determined inference type (step S74). The output means 75 outputs the phrase obtained by the search as the answer to the query (step S75).

[0072] According to the question answering device 70 of the fourth embodiment, it is possible to efficiently generate and output answers to questions based on causal relationship knowledge.

[0073] <Variation> (Variation 1) In the above embodiment, specific semantic relationship pairs are added to causal knowledge, and when inferring an answer to a query, phrases are searched by following causal relationships and specific semantic relationships. Alternatively, the question answering device may search for phrases by following only causal relationships, without considering specific semantic relationships.

[0074] (Modification 2) In the above embodiment, the user's terminal device 5 and the question answering device 100 are connected by communication as shown in Figure 1. However, the question answering device 100 may be configured as an independent device. In that case, the question answering device 100 can be provided with an input unit and a display unit, and the user can input a query into the input unit and the answer to the query can be displayed on the display unit.

[0075] Some or all of the above embodiments may also be described as follows, but are not limited to the following:

[0076] (Note 1) A query retrieval method for obtaining queries, A starting phrase determination means that determines a starting phrase corresponding to the query from multiple phrases included in the causal relationship data, An inference type determination means for determining the inference type based on the aforementioned query, A search means that searches for phrases by tracing the causal relationships contained in the causal relationship data from the starting phrase in a search procedure corresponding to the determined inference type, An output means that outputs the phrase obtained by the search as the answer to the query, A question answering device equipped with the following features.

[0077] (Note 2) The system includes a semantic relationship determination means for extracting specific semantic relationships between multiple phrases included in the aforementioned causal relationship data. The search means is a question answering device as described in Appendix 1, which performs the search by tracing the causal relationships and specific semantic relationships contained in the causal relationship data.

[0078] (Note 3) The question-answering device described in Appendix 2 includes at least one of the following specific semantic relationships: implication relationships, synonymous relationships, and related relationships between phrases.

[0079] (Note 4) The question answering device according to Appendix 1, wherein the starting phrase determination means determines the starting phrase from among a plurality of phrases included in the causal relationship data, the phrase whose meaning is closer to the query phrase than a predetermined criterion.

[0080] (Note 5) The inference type determination means is a question answering device according to Appendix 1 that analyzes the query phrase and determines the inference type corresponding to the inference requested by the phrase.

[0081] (Note 6) The question answering device according to Appendix 1, wherein the inference type includes a first inference type that indicates a search procedure for tracing the causal relationship data in the forward direction of the causal relationship, a second inference type that indicates a search procedure for tracing the causal relationship data in the reverse direction of the causal relationship, and a third inference type that indicates a search procedure that combines the search for tracing the causal relationship data in the forward direction and the search for tracing it in the reverse direction.

[0082] (Note 7) The aforementioned causal relationship data includes data that shows the subclassification of causal relationships. The question answering device described in Appendix 6, which includes a fourth inference type that limits the scope of the search to causal relationships corresponding to predetermined subclassifications among the causal relationship data.

[0083] (Note 8) The output means is a question answering device as described in Appendix 1, which outputs a graph showing the portion of the causal relationship data that was the subject of the search as an answer to the query.

[0084] (Note 9) The aforementioned query is a question concerning the patient's diagnosis, The aforementioned causal relationship data includes causal relationship knowledge related to medical care, constructed from past user medical records and publicly available medical information. The question answering device according to claim 1, wherein the output means outputs an optimized treatment plan as the answer.

[0085] (Note 10) A computer-based question answering method, Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, A question answering method that outputs a phrase obtained through the aforementioned search as an answer to the query.

[0086] (Note 11) Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, A recording medium containing a program that causes a computer to execute a process that outputs the phrase obtained through the search as the answer to the query.

[0087] Although the present disclosure has been described above with reference to embodiments and examples, the present disclosure is not limited to the above embodiments and examples. Various modifications to the structure and details of the present disclosure can be understood by those skilled in the art within the scope of the present disclosure. [Explanation of symbols]

[0088] 5 Terminal devices 13 processors 21, 31 Causal Relationship Database 22, 32 Semantic Relationship Determination Unit 23, 33 Starting Phrase Determination Section 24, 34 Inference Type Determination Unit 25, 35 Reasoning part 36 Relationship Graphs

Claims

1. A query retrieval method for obtaining queries, A starting phrase determination means that determines a starting phrase corresponding to the query from multiple phrases included in the causal relationship data, An inference type determination means for determining the inference type based on the aforementioned query, A search means that searches for phrases by tracing the causal relationships contained in the causal relationship data from the starting phrase in a search procedure corresponding to the determined inference type, An output means that outputs the phrase obtained by the search as the answer to the query, A question answering device equipped with the following features.

2. The system includes a semantic relationship determination means for extracting specific semantic relationships between multiple phrases included in the aforementioned causal relationship data. The question answering device according to claim 1, wherein the search means performs the search by tracing the causal relationships and specific semantic relationships contained in the causal relationship data.

3. The question answering device according to claim 2, wherein the specific semantic relationship includes at least one of an implication relationship, a synonym relationship, and a similar relationship between phrases.

4. The question answering device according to claim 1, wherein the starting phrase determination means determines the starting phrase to be a phrase among a plurality of phrases included in the causal relationship data whose meaning is closer to the query phrase than a predetermined criterion.

5. The question answering device according to claim 1, wherein the inference type determination means analyzes the phrase of the query and determines the inference type corresponding to the inference requested by the phrase.

6. The question answering device according to claim 1, wherein the inference type includes a first inference type that indicates a search procedure for tracing the causal relationship data in the forward direction of the causal relationship, a second inference type that indicates a search procedure for tracing the causal relationship data in the reverse direction of the causal relationship, and a third inference type that indicates a search procedure that combines the search for tracing the causal relationship data in the forward direction and the search for tracing it in the reverse direction.

7. The aforementioned causal relationship data includes data that shows the subclassification of causal relationships. The question answering device according to claim 6, wherein the inference type includes a fourth inference type that limits the search to causal relationships corresponding to predetermined subclassifications among the causal relationship data.

8. The question answering device according to claim 1, wherein the output means outputs a graph showing the portion of the causal relationship data that was the subject of the search as an answer to the query.

9. The aforementioned query is a question concerning the patient's diagnosis, The aforementioned causal relationship data includes causal relationship knowledge related to medical care, constructed from past user medical records and publicly available medical information. The question answering device according to claim 1, wherein the output means outputs an optimized treatment plan as the answer.

10. A computer-based question answering method, Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, A question answering method that outputs a phrase obtained through the aforementioned search as an answer to the query.

11. Get the query, From multiple phrases included in the causal relationship data, determine the starting phrase corresponding to the query. Based on the above query, determine the inference type. In a search procedure corresponding to the determined inference type, the phrase is searched by tracing the causal relationships contained in the causal relationship data from the starting phrase, A program that causes a computer to perform a process that outputs the phrase obtained through the aforementioned search as the answer to the aforementioned query.

Images