Information processing program, information processing device, and information processing method

Abstract

An information processing device (1) receives a hypothesis subject to interpretation; uses a first storage unit to acquire basis information and a rule identifier corresponding to the hypothesis subject to interpretation, said first storage unit including, for each piece of knowledge indicating a plurality of resources and the relationship between the resources, a rule identifier which is associated with a rule used for interpreting a hypothesis and basis information serving as the basis for knowledge; and uses a second storage unit to acquire the probability of being existing knowledge corresponding to the acquired rule identifier, said second storage unit containing, for each rule identifier, the probability that the rule and a hypothesis are existing knowledge. Thus, it is possible to use existing knowledge to interpret a hypothesis in order to screen hypotheses.

Description

[Technical field]

[0001] The present invention relates to an information processing program and the like. [Background technology]

[0002] In recent years, statistically derived hypotheses are used in hypothesis testing for discovery. For example, in genomic medicine, discoveries are made by testing hypotheses about the causal relationship between gene expression or between genes and drug resistance, which are not yet clear.

[0003] Here, a platform has been disclosed that inputs data and extracts results (hypotheses) from statistical causal relationships (see, for example, non-patent literature). It is necessary to select the extracted results (hypotheses) by comparing them with existing knowledge.

[0004] In order to select hypotheses, it is possible to use existing knowledge to interpret the extracted results (hypothesis). For example, such a method can check whether the extracted hypothesis exists in the knowledge data that indicates existing knowledge, and interpret the hypothesis based on the confirmed results. [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] S. Budd et al., "Prototyping CRISP: A Causal Relation and Inference Search Platform applied to Colorectal Cancer Data," 2021 IEEE 3rd Global Conference on Life Sciences and Technologies (LifeTech), 2021, pp. 517-521, doi: 10.1109 / LifeTech52111.2021.9391819. Summary of the Invention [Problem to be solved by the invention]

[0006] However, there is a problem in that a hypothesis cannot be adequately interpreted using existing knowledge. For example, in a method of interpreting a hypothesis using existing knowledge, it is possible to interpret whether a hypothesis is existing knowledge or not, but the interpretation of a hypothesis that is not existing knowledge is the same, and it is not possible to adequately interpret a hypothesis that is not existing knowledge.

[0007] In one aspect, the present invention aims to interpret hypotheses using existing knowledge in order to screen hypotheses. [Means for solving the problem]

[0008] In one aspect, an information processing program causes a computer to execute a process in which, for each piece of knowledge indicating a relationship between a plurality of resources and the resources, a first memory unit including rule identifiers linked to rules used to interpret the hypotheses and basis information that forms the basis of the knowledge is used to obtain basis information and a rule identifier that corresponds to the hypothesis to be interpreted, and, for each rule identifier, a second memory unit including a likelihood that the rule and the hypothesis are existing knowledge is used to obtain a likelihood that the obtained rule identifier is existing knowledge. Effect of the Invention

[0009] According to one embodiment, in order to filter hypotheses, existing knowledge can be used to interpret the hypotheses. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a block diagram illustrating an example of a functional configuration of an information processing device according to a first embodiment. [Diagram 2] FIG. 2 is a diagram for explaining the hypothesis interpretation according to the first embodiment. [Diagram 3] FIG. 3 is a diagram illustrating an example of a hypothesis interpretation DB according to the first embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of an interpretation rule according to the first embodiment. [Diagram 5] FIG. 5 is a diagram illustrating an example of the query information according to the first embodiment. [Figure 6A] FIG. 6A is a diagram (1) showing an example of a hypothesis interpretation according to the first embodiment. [Figure 6B] FIG. 6B is a diagram (2) showing an example of the hypothesis interpretation according to the first embodiment. [Figure 6C] FIG. 6C is a diagram (3) showing an example of the hypothesis interpretation according to the first embodiment. [Figure 6D] FIG. 6D is a diagram (4) showing an example of the hypothesis interpretation according to the first embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of a flowchart of the hypothesis interpretation process according to the first embodiment. [Figure 8] FIG. 8 is a diagram illustrating the effect according to the first embodiment. [Figure 9] FIG. 9 is a block diagram of an example of a functional configuration of an information processing device according to the second embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of knowledge data according to the second embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of interpretation rules according to the second embodiment. [Figure 12] FIG. 12 is a diagram illustrating an example of DB generation according to the second embodiment. [Figure 13] FIG. 13 is a diagram illustrating an example of a flowchart of a DB generation process according to the second embodiment. [Figure 14] FIG. 14 is a block diagram of an example of a functional configuration of an information processing device according to the third embodiment. As illustrated in FIG. [Figure 15A] FIG. 15A is a diagram (1) showing an example of a hypothesis interpretation according to the third embodiment. [Figure 15B] FIG. 15B is a diagram (2) showing an example of the hypothesis interpretation according to the third embodiment. [Figure 15C] FIG. 15C is a diagram (3) showing an example of a hypothesis interpretation according to the third embodiment. [Figure 16A] FIG. 16A is a diagram (1) illustrating an example of a flowchart of a hypothesis interpretation process according to the third embodiment. [Figure 16B]FIG. 16B is a diagram (2) illustrating an example of a flowchart of the hypothesis interpretation process according to the third embodiment. [Figure 17] FIG. 17 is a block diagram of an example of a functional configuration of an information processing device according to the fourth embodiment. As illustrated in FIG. [Figure 18] FIG. 18 is a block diagram of an example of a functional configuration of an information processing device according to the fifth embodiment. As illustrated in FIG. [Figure 19] FIG. 19 is a diagram illustrating an example of generation of interpretation rules according to the fifth embodiment. [Figure 20] FIG. 20 is a diagram illustrating an example of a flowchart of an interpretation rule generation process according to the fifth embodiment. [Figure 21] FIG. 21 is a block diagram of an example of a functional configuration of an information processing device according to the sixth embodiment. As illustrated in FIG. [Figure 22] FIG. 22 is a diagram illustrating an example of a hypothesis interpretation DB according to the sixth embodiment. [Figure 23] FIG. 23 is a diagram showing an example of a screen image. [Figure 24] FIG. 24 is a diagram showing another example of a screen image. [Diagram 25] FIG. 25 is a diagram illustrating an example of a computer that executes an information processing program. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an information processing program, an information processing device, and an information processing method disclosed in the present application will be described in detail with reference to the accompanying drawings. Note that the present invention is not limited to the embodiments. EXAMPLES

[0012] [Functional configuration of information processing device] Fig. 1 is a block diagram showing an example of a functional configuration of an information processing device according to a first embodiment. The information processing device 1 shown in Fig. 1 presents an interpretation of knowledge for a hypothesis by using knowledge data and interpretation rules. The knowledge data here refers to data that holds each piece of knowledge and evidence that is the basis for each piece of knowledge. The interpretation rules here refer to rules for interpreting a hypothesis.

[0013] The information processing device 1 includes a hypothesis interpretation unit 11, a hypothesis interpretation DB 21, interpretation rules 22, and query information 23. The hypothesis interpretation DB 21 is an example of a first storage unit, and the interpretation rules 22 is an example of a second storage unit.

[0014] The hypothesis interpretation DB 21 is a DB used for interpreting hypotheses, and includes a knowledge table 21a and an evidence table 21b.

[0015] The knowledge table 21a stores knowledge according to the structure of the knowledge. The structure of knowledge here is, for example, a structure having one resource, other resources, and their relationships, but is not limited to this and conforms to the structure of knowledge in a knowledge base. However, in the following embodiments, a structure having one resource, other resources, and their relationships is represented as one piece of knowledge, and the structure of a hypothesis is also a similar structure.

[0016] The evidence table 21b includes, for each piece of knowledge, evidence that is the basis of the knowledge and a rule identifier that is linked to a rule used to interpret the hypothesis. That is, the evidence table 21b holds evidence for each piece of knowledge. Here, evidence refers to information that identifies a paper, document, or DB that is evidence. An example of the hypothesis interpretation DB 21 will be described later.

[0017] The interpretation rules 22 include, for each rule identifier, a rule used to interpret a hypothesis and a degree of likelihood that the rule is prior knowledge. An example of the interpretation rules 22 will be described later.

[0018] The query information 23 is a template of a query used when interpreting a hypothesis. An example of the query information 23 will be described later.

[0019] The hypothesis interpretation unit 11 interprets the hypothesis using existing knowledge.

[0020] Here, the hypothesis interpretation according to the first embodiment will be described with reference to FIG. 2. FIG. 2 is a diagram for explaining the hypothesis interpretation according to the first embodiment. As shown in FIG. 2, the hypothesis interpretation unit 11 interprets a hypothesis as to whether it matches existing knowledge, contradicts existing knowledge, or is a new discovery, using a hypothesis interpretation DB 21 and an interpretation rule 22. At this time, the hypothesis interpretation unit 11 calculates the degree of likelihood that it is existing knowledge, and presents it as an interpretation result. For example, the degree of likelihood that it is existing knowledge is "1.0", which means that it matches existing knowledge, and the smaller the value, the more likely it is to be a new discovery.

[0021] Here, an example of the hypothesis interpretation DB 21 will be described with reference to Fig. 3. Fig. 3 is a diagram showing an example of the hypothesis interpretation DB according to the embodiment 1. The hypothesis interpretation DB 21 has a knowledge table 21a and an evidence table 21b.

[0022] The knowledge table 21a stores gene IDs (IDentifiers), relationships, and gene IDs in association with knowledge IDs (IDentifiers). A gene ID, a relationship, and a gene ID are one piece of knowledge. As an example, when the knowledge ID is "0", one gene ID is stored as "0", the other gene ID is stored as "1", and the relationship is stored as "increase". In other words, knowledge is stored that indicates a relationship in which when the gene indicated by one gene ID "0" increases, the gene indicated by the other gene ID "1" also increases.

[0023] The evidence table 21b stores evidence in association with a knowledge ID and a matching rule ID. For knowledge, the evidence stores an identifier for identifying the paper, document, or DB that serves as evidence. The knowledge ID corresponds to the knowledge ID in the knowledge table 21a. The matching rule ID corresponds to the rule ID of the interpretation rule 22, which will be described later. As an example, when the knowledge ID is "0" and the matching rule ID is "0", "pmidxxxx" is stored as evidence.

[0024] An example of the interpretation rule 22 will now be described with reference to Fig. 4. Fig. 4 is a diagram illustrating an example of the interpretation rule according to the first embodiment.

[0025] The interpretation rules 22 store relationships, rule contents, and scores in association with rule IDs. Relationships correspond to relationships between resources. Rule contents indicate the content of rules expressed from resources and relationships. Scores indicate the degree of certainty that the knowledge is existing knowledge. In other words, scores are the degree of match with existing knowledge. A larger score value means that the knowledge is more likely to be existing knowledge. A smaller score value means that the knowledge is more likely to be an appropriate new discovery. As an example, scores are expressed as numbers between 0 and 1.

[0026] For example, when the rule ID is "0", the relationship is "increase", the rule content is "the same gene has an 'increase' relationship", and the score is "1". When the rule ID is "2", the relationship is "increase", the rule content is "the same gene has an 'increase' relationship", and the score is "0.8".

[0027] Returning to FIG. 1, for example, when the hypothesis interpretation unit 11 acquires a hypothesis to be interpreted, it acquires a knowledge ID corresponding to the hypothesis from the knowledge table 21a of the hypothesis interpretation DB 21. Then, the hypothesis interpretation unit 11 acquires a list of evidence and rules linked to the acquired knowledge ID from the evidence table 21b of the hypothesis interpretation DB 21 and the interpretation rule 22. Then, the hypothesis interpretation unit 11 acquires a score of the corresponding rule from the acquired list of rules. Then, the hypothesis interpretation unit 11 holds information associating the knowledge ID, evidence, matched rule ID, and score as a match result list. Note that the hypothesis interpretation unit 11 may execute the process from acquiring the hypothesis to be interpreted to holding the match result list by using a query. An example of a query will be described later.

[0028] Then, the hypothesis interpretation unit 11 aggregates the held match result list. Note that, when aggregating, the hypothesis interpretation unit 11 may, for example, give priority to information with a high score included in the match result list, but as another example, the scores may be averaged and all evidence may be listed together, and the present invention is not limited to this.

[0029] Here, an example of the query information will be described with reference to FIG. 5. FIG. 5 is a diagram showing an example of the query information according to the first embodiment. As shown in FIG. 5, the query information 23 is information that associates a hypothesis structure, a matching knowledge acquisition query template, and a contradiction knowledge acquisition query template. The hypothesis structure corresponds to a relationship between resources indicated in the hypothesis. The matching knowledge acquisition query template is a query template for acquiring knowledge that matches the hypothesis structure. The contradiction knowledge acquisition query template is a query template for acquiring contradictory knowledge that does not match the hypothesis structure.

[0030] As an example, when the hypothesis structure is "increase", "SELECT...relation=increase" is stored as the matching knowledge acquisition query template, and "SELECT...relation=decrease" is stored as the contradictory knowledge acquisition query template.

[0031] 6A to 6D are diagrams showing an example of a hypothesis interpretation according to Example 1. As shown in Fig. 6A, it is assumed that the hypothesis interpretation unit 11 acquires "A-(increase)→B" (a relationship in which B increases as A increases) as a hypothesis to be interpreted. As an example, the hypothesis has the content shown by the symbol h0, and the hypothesis structure is "increase".

[0032] Then, the hypothesis interpretation unit 11 acquires a matching knowledge acquisition query template and a contradiction knowledge acquisition query template corresponding to the hypothesis structure of the hypothesis h0 acquired from the query information 23. Here, a matching knowledge acquisition query template and a contradiction knowledge acquisition query template whose hypothesis structure is “increase” are acquired from the query information 23.

[0033] Then, the hypothesis interpretation unit 11 generates a matching knowledge acquisition query and a contradiction knowledge acquisition query from the hypothesis h0, the matching knowledge acquisition query template and the contradiction knowledge acquisition query template. Here, the matching knowledge acquisition query and the contradiction knowledge acquisition query are generated by substituting "A" and "B" for the "$(Gene1)" and "$(Gene2)" parts of the matching knowledge acquisition query template and the contradiction knowledge acquisition query template, respectively. The left diagram at the bottom of Fig. 6A is the generated matching knowledge acquisition query, and the right diagram at the bottom of Fig. 6A is the generated contradiction knowledge acquisition query.

[0034] Then, as shown in FIG. 6B, the hypothesis interpretation unit 11 queries the hypothesis interpretation DB 21 and the interpretation rule 22 using the matched knowledge acquisition query. The hypothesis interpretation unit 11 acquires a match result list. Here, from the knowledge table 21a, a gene ID of "A", a gene ID of "B", and a knowledge ID of "0" for which the relationship is "increased" are acquired (symbol a1). From the evidence table 21b, a match rule ID of "1" and evidence of "pmid1234" for which the knowledge ID is "0", another match rule ID of "2" and evidence of "pmid5678" are acquired (symbol a2). In addition, from the interpretation rule 22, a score of "0.8" for which the match rule ID is "1" and a score of "0.6" for which the match rule ID is "2" are acquired (symbol a3). Then, the hypothesis interpretation unit 11 holds information in which the knowledge ID, the evidence, the match rule ID, and the score are associated with each other as a match result list. Here, the match result list is the information indicated by the symbol a4.

[0035] Then, as shown in FIG. 6C, the hypothesis interpretation unit 11 queries the hypothesis interpretation DB 21 and the interpretation rule 22 using the contradiction knowledge acquisition query. The hypothesis interpretation unit 11 acquires a contradiction result list. Here, from the knowledge table 21a, a gene ID of "A", a gene ID of "B", and a knowledge ID of "1" with a relationship of "decreasing" are acquired (symbol b1). From the evidence table 21b, a matched rule ID of "3" with a knowledge ID of "1" and "pmid910" with evidence are acquired (symbol b2). In addition, from the interpretation rule 22, a score of "0.6" with a matched rule ID of "3" is acquired (symbol b3). Then, the hypothesis interpretation unit 11 holds information in which the knowledge ID, the evidence, the matched rule ID, and the score are associated with each other as a contradiction result list. Here, the contradiction result list is information indicated by symbol b4.

[0036] Then, as shown in Fig. 6D, the hypothesis interpretation unit 11 aggregates the held matching result list and contradictory result list and holds them as an aggregated result. Here, the aggregated results are held in the order of the matching result list, then the contradictory result list, in descending order of score. Here, the aggregated results of the knowledge ID, evidence, and matching rule ID are held in the order of scores "0.8", "0.6", and "-0.6". The reason the last score in the aggregated result is negative is because it is the score for the contradictory knowledge.

[0037] FIG. 7 is a diagram illustrating an example of a flowchart of the hypothesis interpretation process according to the first embodiment.

[0038] 7, the hypothesis interpretation unit 11 acquires a hypothesis H, a matching knowledge acquisition query template Q_mt, and a contradiction knowledge acquisition query template Q_ct (step S11). For example, when the hypothesis interpretation unit 11 acquires a hypothesis H to be interpreted, it acquires, from the query information 23, a matching knowledge acquisition query template Q_mt and a contradiction knowledge acquisition query template Q_ct corresponding to a relationship included in the hypothesis H.

[0039] Then, the hypothesis interpretation unit 11 generates a matching knowledge acquisition query Q_m and a contradiction knowledge acquisition query Q_c from the hypothesis H, the matching knowledge acquisition query template Q_mt and the contradiction knowledge acquisition query template Q_ct (step S12). For example, the hypothesis interpretation unit 11 substitutes the gene IDa and the gene IDb included in the hypothesis H into the matching knowledge acquisition query template Q_mt and the contradiction knowledge acquisition query template Q_ct to generate the matching knowledge acquisition query Q_m and the contradiction knowledge acquisition query Q_c.

[0040] The hypothesis interpretation unit 11 then queries the hypothesis interpretation DB (21) / interpretation rules (22) using the matching knowledge acquisition query Q_m and the contradiction knowledge acquisition query Q_c, and obtains a matching result list R_m and a contradiction result list R_c as query results (step S13).The hypothesis interpretation unit 11 then generates an aggregated result R from the matching result list R_m and the contradiction result list R_c (step S14).

[0041] Then, the hypothesis interpretation unit 11 outputs the aggregation result R (step S15). Then, the hypothesis interpretation unit 11 ends.

[0042] [Effects of Example 1] Here, the effect of the first embodiment will be described with reference to FIG. 8. FIG. 8 is a diagram showing the effect of the first embodiment. As shown in FIG. 8, the hypothesis interpretation unit 11 interprets the hypothesis to be interpreted using the hypothesis interpretation DB 21 and the interpretation rule 22, and presents evidence and a score as the interpretation result. This allows the hypothesis interpretation unit 11 to provide an interpretation for the hypothesis to be interpreted. As a result, the hypothesis interpretation unit 11 becomes able to select the hypothesis to be interpreted. In addition, the hypothesis interpretation unit 11 becomes able to customize and improve the interpretation rule 22 by allowing the presented evidence and score to be confirmed.

[0043] Thus, according to the first embodiment, the information processing device 1 receives a hypothesis to be interpreted. The information processing device 1 acquires basis information and a rule identifier corresponding to the hypothesis to be interpreted, using a hypothesis interpretation DB 21 including basis information that is the basis of knowledge for each piece of knowledge indicating a relationship between a plurality of resources and the resources, and a rule identifier associated with a rule used to interpret the hypothesis. Then, the information processing device 1 acquires, for each rule identifier, a degree of likelihood that the rule and the hypothesis are existing knowledge, using an interpretation rule 22 including a degree of likelihood that the rule and the hypothesis are existing knowledge. This allows the information processing device 1 to interpret the hypothesis to be interpreted using existing knowledge in order to select the hypothesis to be interpreted.

[0044] According to the first embodiment, the information processing device 1 presents the acquired evidence information for the hypothesis to be interpreted and the degree of likelihood that the hypothesis is existing knowledge. This allows the information processing device 1 to select the hypothesis to be interpreted. The hypothesis interpretation unit 11 also allows the interpretation rule 22 to be customized and improved. EXAMPLES

[0045] Incidentally, it has been explained that the information processing device 1 according to the first embodiment interprets a hypothesis using the hypothesis interpretation DB 21 and the interpretation rule 22, and presents evidence and a score as an interpretation result. The hypothesis interpretation DB 21 used for the interpretation is generated in advance. However, the hypothesis interpretation DB 21 used for the interpretation may be generated using knowledge data between any resources.

[0046] In the second embodiment, a case will be described in which the hypothesis interpretation DB 21 used for interpretation is generated using knowledge data between arbitrary resources.

[0047] [Functional configuration of information processing device] Fig. 9 is a block diagram showing an example of a functional configuration of an information processing device according to a second embodiment. Note that, in the functional configuration of the information processing device 1 according to the second embodiment, the same components as those of the information processing device 1 shown in Fig. 1 are indicated with the same reference numerals, and the description of the overlapping configurations and operations will be omitted. The first embodiment differs from the second embodiment in that knowledge data 24 and a DB generating unit 12 are added. The first embodiment differs from the second embodiment in that the interpretation rule 22 is changed to an interpretation rule 22A. Note that the knowledge data 24 is an example of a third storage unit.

[0048] The knowledge data 24 is a knowledge base that indicates knowledge between any resources. The knowledge data 24 stores a plurality of pieces of data each of which is a single piece of knowledge consisting of a subject, an object, and a relationship (predicate) between them. In addition, an evidence resource is associated with each piece of knowledge.

[0049] Here, an example of the knowledge data 24 will be described with reference to FIG. 10. FIG. 10 is a diagram showing an example of knowledge data according to the second embodiment. As shown in FIG. 10, the knowledge data stores a subject ID, a predicate ID, and an object ID in association with a knowledge ID. Furthermore, the knowledge data includes evidence resource data. The evidence resource data stores a knowledge-evidence set ID and an evidence resource in association with a knowledge ID. The subject ID and the object ID include, for example, a gene ID. The predicate ID includes a relationship between genes. The evidence resource is an identifier for identifying a paper, a document, or a DB that is evidence of knowledge. Although the knowledge data 24 is expressed in a table structure, the present invention is not limited to this and may be expressed in a graph structure using RDF (Resource Description Framework) or the like, and the evidence resource data may also be stored in association with any substructure of the graph structure.

[0050] As an example, when the knowledge ID is "0", "0" is stored as the subject ID, "increase" as the predicate ID, and "1" as the object ID. Also, when the knowledge ID is "0", two evidence resources are stored. For one evidence resource, "0" is stored as the knowledge-evidence pair ID, and "pmid1234" is stored as the evidence resource. For the other evidence resource, "1" is stored as the knowledge-evidence pair ID, and "pmid567" is stored as the evidence resource.

[0051] Returning to FIG. 9, the interpretation rules 22A include, for each rule identifier, an interpretation rule for interpreting a hypothesis, a degree of certainty that the rule is existing knowledge, and a query template to be linked to the knowledge data 24.

[0052] Here, an example of the interpretation rule 22A will be described with reference to FIG. 11. FIG. 11 is a diagram showing an example of the interpretation rule according to the second embodiment. Regarding the data configuration of the interpretation rule 22A according to the second embodiment, the same data configuration as that of the interpretation rule 22 shown in FIG. 4 will not be described. The interpretation rule 22A differs from the interpretation rule 22 in that a rule query template and an evidence query template are added. The rule query template is a query template that links a rule with a knowledge ID of the knowledge data 24. The evidence query template is a query template that links a rule with an evidence resource included in the knowledge data 24. The rule query template and the evidence query template are defined according to the structure of the knowledge data and the type of DB. For example, when the knowledge data is expressed in a graph structure by RDF, the rule query template and the evidence query template may be written in a format such as SPARQL, which is a query language for RDF.

[0053] As an example, when the rule ID is "0", the relationship is "increase", the rule content is "the same gene has an 'increase' relationship", and the score is "1". In addition, a rule query template and an evidence query template are stored.

[0054] Returning to FIG. 9, the DB generating unit 12 generates the hypothesis interpretation DB 21 from the knowledge data 24 and the interpretation rule 22A. For example, the DB generating unit 12 acquires a resource set that is the subject of a hypothesis. The DB generating unit 12 performs the following process for each rule of the interpretation rule 22A for the acquired resource set. The DB generating unit 12 generates a query from the resource set and a rule query template corresponding to one rule ID of the interpretation rule 22A. Then, the DB generating unit 12 queries the knowledge data 24 with the generated query to acquire a query result as to whether or not a knowledge ID that matches the rule exists. If the query result indicates that a knowledge ID that matches the rule exists, the DB generating unit 12 generates a query from the knowledge ID and an evidence query template corresponding to the same rule ID. Then, the DB generating unit 12 queries the knowledge data 24 with the generated query to acquire an evidence resource corresponding to the knowledge indicated by the knowledge ID. Then, the DB generating unit 12 links the resource set with the relationship included in the rule, and updates the knowledge table 21a of the hypothesis interpretation DB 21. When updating, a knowledge ID is assigned. In addition, the DB generation unit 12 links the evidence resource to the rule ID and the assigned knowledge ID, and updates the evidence table 21b of the hypothesis interpretation DB 21. In this way, the DB generation unit 12 performs DB generation processing for each resource set in the list of resource sets that are the subject of hypotheses.

[0055] Fig. 12 is a diagram illustrating an example of DB generation according to Example 2. As illustrated in Fig. 12, it is assumed that the DB generating unit 12 acquires "A, B" as one resource pair.

[0056] The DB generation unit 12 selects one rule from the interpretation rules 22A. Here, the rule with the rule ID "0" is selected (symbol c1). Then, the DB generation unit 12 determines whether the resource set matches the rule. That is, the DB generation unit 12 generates a query from the rule query template for the rule and the resource set. Here, the query is generated by substituting "A" and "B" for the "$(Gene1)" and "$(Gene2)" parts of the rule query template.

[0057] Then, the DB generating unit 12 queries the knowledge data 24 with the generated query, and obtains a query result indicating whether or not a knowledge ID that matches the rule exists. Here, a query result indicating the knowledge ID "1234" is obtained. That is, the resource set matches the rule, and a knowledge ID that matches the rule exists.

[0058] Then, if the query result indicates that a knowledge ID matching the rule exists, the DB generation unit 12 generates an evidence acquisition query from the knowledge ID and an evidence query template corresponding to the same rule ID. Here, the evidence acquisition query is generated by substituting "1234" into the "knowledge ID" portion of the evidence query template.

[0059] Then, the DB creation unit 12 queries the knowledge database 24 with the created evidence acquisition query to acquire an evidence resource corresponding to the knowledge indicated by the knowledge ID. In this example, "pmid789" is acquired as the evidence resource.

[0060] Then, the DB generating unit 12 links the relationship included in the rule and the newly assigned knowledge ID to the resource set, and updates the knowledge table 21a of the hypothesis interpretation DB 21. Here, when the knowledge ID is "0", the knowledge table 21a is updated to "A" as the gene ID, "increase" as the relationship, and "B" as the gene ID.

[0061] In addition, the DB creation unit 12 links the evidence resource to the rule ID and the assigned knowledge ID, and updates the evidence table 21b of the hypothesis interpretation DB 21. Here, in the evidence table 21b, when the knowledge ID is "0" and the matching rule ID is "0", "pmid789" is updated as evidence.

[0062] FIG. 13 is a diagram illustrating an example of a flowchart of a DB generation process according to the second embodiment.

[0063] 13, the DB generating unit 12 acquires a list S_list of target resource sets (step S21). Note that the list of target resource sets may be, for example, a list generated in advance, or a list generated from hypotheses extracted in some hypothesis extraction phase.

[0064] Then, the DB generating unit 12 acquires the rule list R_list from the interpretation rule 22A (step S22). The DB generating unit 12 judges whether or not there is an unselected resource set in the resource set list S_list (step S23). If it is judged that there is an unselected resource set in the resource set list S_list (step S23; Yes), the DB generating unit 12 selects a resource set S from the resource set list S_list (step S24).

[0065] Then, the DB generating unit 12 judges whether or not there is an unselected rule in the rule list R_list (step S25). If it is judged that there is no unselected rule in the rule list R_list (step S25; No), the DB generating unit 12 proceeds to step S23 to select the next resource set.

[0066] On the other hand, if it is determined that an unselected rule exists in the rule list R_list (step S25; Yes), the DB generation unit 12 selects a rule R from the rule list R_list (step S26). The DB generation unit 12 generates a query Q from the query template of the rule R and the resource set S, queries the knowledge data 24 with the query Q, and obtains a query result R_match (step S27). Then, the DB generation unit 12 determines whether the number of query results R_match is 1 or more (step S28).

[0067] If it is determined that the number of query results R_match is not 1 or more (step S28; No), the DB generating unit 12 proceeds to step S25 to select the next rule.

[0068] On the other hand, if it is determined that the number of query results R_match is one or more (Step S28; Yes), the DB generation unit 12 generates an evidence acquisition query Q_evi from the query template of the rule R, the resource set S and the query result R_match, queries the knowledge data 24 with the evidence acquisition query Q_evi, and obtains the query result R_evi (Step S29).

[0069] Then, the DB generating unit 12 updates the hypothesis interpretation DB 21 based on the query results R_match and R_evi (step S30). That is, the DB generating unit 12 updates the knowledge table 21a and the evidence table 21b of the hypothesis interpretation DB 21. Then, the DB generating unit 12 proceeds to step S25 to select the next rule.

[0070] In step S23, if the DB generating unit 12 determines that there is no unselected resource set in the list S_list of resource sets (step S23; No), the DB generating unit 12 ends the DB generation process.

[0071] [Effects of Example 2] Thus, according to the second embodiment, the information processing device 1 further receives a set of resources, and acquires basis information from the knowledge data 24 that links, for each piece of knowledge including a subject, an object, and a predicate indicating a relationship between the subject and the object, basis information that serves as the basis of the knowledge, based on the rules included in the interpretation rules 22A. Then, the information processing device 1 adds, for the set of resources, information including the relationship between the resources obtained from the rules of the interpretation rules 22A, the basis information, and a rule identifier linked to the rule, to the hypothesis interpretation DB 21. As a result, the information processing device 1 can generate the hypothesis interpretation DB 21 used when interpreting a hypothesis by using the knowledge data 24 and the interpretation rules 22A. EXAMPLES

[0072] Incidentally, it has been explained that the information processing device 1 according to the first embodiment interprets a hypothesis using the hypothesis interpretation DB 21 and the interpretation rules 22, and presents evidence and a score as an interpretation result. Also, it has been explained that the information processing device 1 according to the second embodiment generates the hypothesis interpretation DB 21 from the knowledge data 24 and the interpretation rules 22A. However, the information processing device 1 is not limited to this, and may interpret a hypothesis using the interpretation rules 22A and the knowledge data 24 instead of the hypothesis interpretation DB 21, and present evidence and a score as an interpretation result.

[0073] In the third embodiment, an information processing device 1 that interprets a hypothesis using knowledge data 24 and interpretation rules 22A and presents evidence and a score as the interpretation result will be described.

[0074] [Functional configuration of information processing device] Fig. 14 is a block diagram showing an example of a functional configuration of an information processing device according to Example 3. Note that, in the functional configuration of the information processing device 1 according to Example 3, the same components as those of the information processing device 1 shown in Fig. 9 are indicated with the same reference numerals, and the description of the overlapping configurations and operations will be omitted. Example 2 differs from Example 3 in that the hypothesis interpretation DB 21 and the DB generating unit 12 are deleted. Also, Example 2 differs from Example 3 in that the hypothesis interpretation unit 11 is changed to a hypothesis interpretation unit 11A.

[0075] The hypothesis interpretation unit 11A interprets a hypothesis using the knowledge data 24 and the interpretation rule 22A. For example, when the hypothesis interpretation unit 11A acquires a hypothesis to be interpreted, the following process is performed for each rule of the interpretation rule 22A for the acquired hypothesis to be interpreted. The hypothesis interpretation unit 11A generates a query from the hypothesis and a rule query template corresponding to one rule ID of the interpretation rule 22A. Then, the hypothesis interpretation unit 11A queries the knowledge data 24 with the generated query to acquire a query result as to whether or not a knowledge ID matching the rule exists. If the query result indicates that a knowledge ID matching the rule exists, the hypothesis interpretation unit 11A generates a query from the knowledge ID and an evidence query template corresponding to the same rule ID. Then, the hypothesis interpretation unit 11A queries the knowledge data 24 with the generated query to acquire an evidence resource corresponding to the knowledge indicated by the knowledge ID. Then, the hypothesis interpretation unit 11A holds information in which the knowledge ID, evidence, matching rule ID, and score are associated with each other as a matching result list.

[0076] Then, the hypothesis interpretation unit 11A aggregates the held match result list. Note that, when aggregating, the hypothesis interpretation unit 11A may, for example, prioritize information included in the match result list that has a high score, but as another example, the scores may be averaged and all evidence may be listed together, and the present invention is not limited to this.

[0077] 15A to 15C are diagrams showing an example of a hypothesis interpretation according to Example 3. As shown in Fig. 15A, it is assumed that the hypothesis interpretation unit 11A acquires a hypothesis "A-(increase)→B" (a relationship in which B increases as A increases). As an example, the hypothesis has the content shown by the symbol h0, and the hypothesis structure is "increase".

[0078] The hypothesis interpretation unit 11A selects one rule from the interpretation rules 22A. Here, the rule with the rule ID "0" is selected (symbol d1). Then, the hypothesis interpretation unit 11A determines whether the hypothesis matches the rule. That is, the hypothesis interpretation unit 11A generates a query from the rule query template for the rule and the hypothesis. Here, the query is generated by substituting "A" and "B" for the "$(Gene1)" and "$(Gene2)" parts of the rule query template.

[0079] Then, the hypothesis interpretation unit 11A queries the knowledge data 24 with the generated query, and obtains a query result indicating whether or not a knowledge ID that matches the rule exists. Here, a query result indicating the knowledge ID "1234" is obtained. In other words, the hypothesis matches the rule, and a knowledge ID that matches the rule exists.

[0080] Then, when the query result indicates that a knowledge ID that matches the rule exists, the hypothesis interpretation unit 11A generates an evidence acquisition query from the knowledge ID and an evidence query template that corresponds to the same rule ID. Here, the evidence acquisition query is generated by substituting "1234" into the "knowledge ID" portion of the evidence query template.

[0081] Then, the hypothesis interpretation section 11A queries the knowledge data 24 with the generated evidence acquisition query to acquire an evidence resource corresponding to the knowledge indicated by the knowledge ID. In this example, "pmid789" is acquired as the evidence resource.

[0082] The hypothesis interpretation unit 11A then holds information in which the knowledge ID, the evidence, the matching rule ID, and the score are associated with each other as a matching result list. Here, the matching result list is information indicated by the reference symbol d3.

[0083] Next, as shown in Fig. 15B, the hypothesis interpretation unit 11A receives a contradictory hypothesis "A-(decrease) → B" (a relationship in which B decreases when A decreases) for "A-(increase) → B" acquired as a hypothesis. As an example, the contradictory hypothesis is the content shown by symbol h1, and the hypothesis structure is "decrease" which contradicts the hypothesis.

[0084] The hypothesis interpretation unit 11A selects one rule from the interpretation rules 22A. Here, the rule with the rule ID "1" is selected (symbol d2). Then, the hypothesis interpretation unit 11A determines whether the hypothesis matches the rule. That is, the hypothesis interpretation unit 11A generates a query from the rule query template for the rule and the hypothesis. Here, the query is generated by substituting "A" and "B" for the "$(Gene1)" and "$(Gene2)" parts of the rule query template.

[0085] Then, the hypothesis interpretation unit 11A queries the knowledge data 24 with the generated query, and obtains a query result indicating whether or not a knowledge ID that matches the rule exists. Here, a query result indicating the knowledge ID "5678" is obtained. In other words, the contradictory hypothesis matches the rule, and a knowledge ID that matches the rule exists.

[0086] Then, when the query result indicates that a knowledge ID that matches the rule exists, the hypothesis interpretation unit 11A generates an evidence acquisition query from the knowledge ID and an evidence query template that corresponds to the same rule ID. Here, the evidence acquisition query is generated by substituting "5678" into the "knowledge ID" portion of the evidence query template.

[0087] Then, the hypothesis interpretation section 11A queries the knowledge data 24 with the generated evidence acquisition query to acquire an evidence resource corresponding to the knowledge indicated by the knowledge ID. In this example, "pmid1234" is acquired as the evidence resource.

[0088] The hypothesis interpretation unit 11A then holds information in which the knowledge ID, the evidence, the matching rule ID, and the score are associated with each other as a contradiction result list. Here, the contradiction result list is information indicated by the reference symbol d4.

[0089] Then, as shown in Fig. 15C, hypothesis interpretation unit 11A aggregates the held match result list and contradiction result list and holds them as an aggregated result. Here, the aggregated results are held in the order of match result list, contradiction result list, and in descending order of score. Here, the aggregated results of knowledge ID, evidence, and matched rule ID are held in the order of scores "1", "-1". The reason the last score in the aggregated result is negative is because it is the score for the contradiction hypothesis.

[0090] 16A and 16B are diagrams illustrating an example of a flowchart of a hypothesis interpretation process according to the third embodiment.

[0091] 16A, the hypothesis interpretation unit 11A acquires a hypothesis H (step S41). Then, the hypothesis interpretation unit 11A inputs the hypothesis H and calls a process for generating a match result list R_m (step S42). The process for generating a result list from a hypothesis will be described later.

[0092] Next, the hypothesis interpretation unit 11A generates a contradictory hypothesis H_c from the hypothesis H (step S43). Then, the hypothesis interpretation unit 11A inputs the contradictory hypothesis Hc and calls a process for generating a contradictory result list R_c (step S44). The process for generating a result list from a hypothesis will be described later.

[0093] Then, the hypothesis interpretation section 11A generates the aggregated result R_aggr from the matched result list R_m and the contradictory result list R_c (step S45).

[0094] Then, the hypothesis interpretation section 11A outputs the aggregation result R_aggr (step S46), and the hypothesis interpretation section 11A ends.

[0095] 16B, the hypothesis interpretation unit 11A executes a process of generating a result list of matches or contradictions as follows: When a hypothesis H is input, the hypothesis interpretation unit 11A acquires a rule list R_list related to the hypothesis H from the interpretation rules 22A (step S51). For example, when the hypothesis is "A-(increase)→B", the hypothesis interpretation unit 11A acquires a rule list indicating an increase in the relationship from the interpretation rules 22A.

[0096] The hypothesis interpretation unit 11A judges whether or not an unselected rule exists in the rule list R_list (step S52). If it is judged that an unselected rule exists in the rule list R_list (step S52; Yes), the hypothesis interpretation unit 11A selects a rule R from the rule list R_list (step S53).

[0097] Then, the hypothesis interpretation unit 11A judges whether or not the judgment based on the rule R can be omitted based on the aggregation method and the rule list R_list (step S54). If it is judged that the judgment based on the rule R can be omitted (step S54; Yes), the hypothesis interpretation unit 11A proceeds to step S52 to select the next rule.

[0098] On the other hand, if it is determined that the judgment based on the rule R cannot be omitted (step S54; No), the hypothesis interpretation unit 11A generates a query Q from the rule query template of the rule R and the input hypothesis H. Then, the hypothesis interpretation unit 11A queries the knowledge data 24 with the generated query Q and obtains the query result R_match (step S55).

[0099] Then, the hypothesis interpretation unit 11A judges whether or not the number of the query results R_match is 1 or more (step S56). If it is judged that the number of the query results R_match is not 1 or more (step S56; No), the hypothesis interpretation unit 11A proceeds to step S52 to select the next rule.

[0100] On the other hand, if it is determined that the number of query results R_match is 1 or more (step S56; Yes), the hypothesis interpretation unit 11A generates an evidence acquisition query Q_evi from the evidence query template of the rule R, the input hypothesis H, and the query result R_match. Then, the hypothesis interpretation unit 11A queries the knowledge data 24 for the evidence acquisition query Q_evi and acquires the query result R_evi (step S57).

[0101] Then, the hypothesis interpretation section 11A adds the set of the rule R, the query result R_match, and the query result R_evi to the result list R_list (step S58).Then, the hypothesis interpretation section 11A proceeds to step S52 to select the next rule.

[0102] In step S52, if the hypothesis interpretation section 11A determines that there is no unselected rule in the rule list R_list (step S52; No), the hypothesis interpretation section 11A returns the result list R_list to the caller.

[0103] [Effects of Example 3] Thus, according to the third embodiment, the information processing device 1 receives a hypothesis to be interpreted, and uses a rule related to the hypothesis to be interpreted, which is included in the interpretation rules 22A including the rule used to interpret the hypothesis and the likelihood that the hypothesis is existing knowledge for each rule identifier, to determine whether or not knowledge corresponding to the hypothesis to be interpreted exists in the knowledge data 24 in which grounds information that is the grounds for knowledge is linked to each piece of knowledge including a subject, an object, and a predicate indicating the relationship between the subject and the object. Then, when the information processing device 1 determines that knowledge corresponding to the hypothesis to be interpreted exists in the knowledge data 24, it acquires the grounds information corresponding to the knowledge and the likelihood that the knowledge is existing knowledge corresponding to the rule related to the hypothesis to be interpreted. As a result, the information processing device 1 can interpret the hypothesis to be interpreted using the knowledge data 24 and the interpretation rules 22A without generating the hypothesis interpretation DB 21, and present evidence and a score as an interpretation result. EXAMPLES

[0104] Incidentally, it has been explained that the information processing device 1 according to the first embodiment interprets a hypothesis using the hypothesis interpretation DB 21 and the interpretation rule 22, and presents evidence and a score as an interpretation result. It has been explained that the information processing device 1 according to the second embodiment generates the hypothesis interpretation DB 21 used for interpretation using the knowledge data 24 and the interpretation rule 22A. Here, the knowledge data 24 is a knowledge base indicating existing knowledge between any resources. Such existing knowledge exists in various forms such as papers, documents, DBs, etc., and includes unstructured data and data that is not machine-readable. Therefore, the information processing device 1 may generate the knowledge data 24 by utilizing a relationship extraction process from natural language for papers, documents, etc., and by utilizing a process for DB integration for DBs, etc.

[0105] In the fourth embodiment, an information processing device 1 that generates knowledge data 24 by utilizing a process for extracting relationships from a natural language and a process for integrating DBs will be described.

[0106] [Functional configuration of information processing device] Fig. 17 is a block diagram showing an example of a functional configuration of an information processing device according to Example 4. Note that, regarding the functional configuration of the information processing device 1 according to Example 4, the same components as those of the information processing device 1 shown in Fig. 9 are indicated with the same reference numerals, and the description of the overlapping configurations and operations will be omitted. Example 4 is different from Example 2 in that a relationship extraction unit 31 and a DB integration unit 32 are added.

[0107] The relationship extraction unit 31 extracts relationships from natural language. For example, rules for relationship extraction are defined in advance. The relationship extraction unit 31 determines whether the target paper and document match the predefined rules for relationship extraction. If the target paper and document match the rules for relationship extraction, the relationship extraction unit 31 generates relationship data using the paper and document that match the rules as evidence resources. The relationship data is one piece of knowledge. The relationship extraction unit 31 links the evidence resource to one piece of knowledge and stores it in the knowledge data 24. Note that the method for realizing the relationship extraction is not limited, and machine learning, a language model, or any existing technology may be used.

[0108] The DB integrating unit 32 integrates multiple existing DBs. For example, a rule for DB integration is defined in advance. The DB integration rule may be, for example, a rule for integrating entities or a rule for unifying items. The DB integrating unit 32 integrates multiple existing target DBs according to the predefined DB integration rule. Then, the DB integrating unit 32 generates knowledge between resources using the integrated DB as an evidence resource. Then, the DB integrating unit 32 links the evidence resource to one piece of knowledge and stores it in the knowledge data 24.

[0109] [Effects of Example 4] Thus, according to the fourth embodiment, the information processing device 1 uses a predetermined relationship extraction process from natural language to generate knowledge data 24 with the target document as basis information. Then, the information processing device 1 uses a predetermined integration process for integrating multiple existing DBs to integrate the multiple target DBs and generate knowledge data 24 with the integrated DB as basis information. This enables the information processing device 1 to generate knowledge data 24 from documents and DBs that exist in different formats. EXAMPLES

[0110] Incidentally, the information processing device 1 according to the first embodiment has been described as interpreting a hypothesis using the hypothesis interpretation DB 21 and the interpretation rules 22, and presenting evidence and a score as an interpretation result. The information processing device 1 according to the second embodiment has been described as generating the hypothesis interpretation DB 21 used for interpretation using the knowledge data 24 and the interpretation rules 22A. The information processing device 1 according to the fourth embodiment has been described as generating the knowledge data 24 by utilizing a relationship extraction process from natural language and a process for DB integration. Here, the interpretation rules 22 and 22A used for interpretation are often created by humans. When rules are created by humans, there is a problem that it takes time, for example.

[0111] In the fifth embodiment, an information processing device 1 that generates the interpretation rule 22A will be described.

[0112] [Functional configuration of information processing device] Fig. 18 is a block diagram showing an example of a functional configuration of an information processing device according to Example 5. Note that, regarding the functional configuration of the information processing device 1 according to Example 5, the same components as those of the information processing device 1 shown in Fig. 17 are indicated with the same reference numerals, and the description of the overlapping configurations and operations will be omitted. Example 5 is different from Example 4 in that an interpretation rule generating unit 33 is added.

[0113] The interpretation rule generation unit 33 generates the interpretation rule 22A. For example, the interpretation rule generation unit 33 acquires a list of resource pairs and specifies the relationships between the resources to be extracted. The interpretation rule generation unit 33 extracts, for example, from the knowledge data 24, a feature table corresponding to the resource pairs included in the list and the specified relationships. As an example, when the knowledge data 24 is a graph DB, the interpretation rule generation unit 33 can extract features from a graph pattern of a graph structure between entities.

[0114] The interpretation rule generation unit 33 then applies explainable AI to the feature table as input and the variable indicating the presence or absence of a specified relationship as the objective variable, to calculate a list of scores for each feature included in the feature table. The explainable AI (Artificial Intelligence) referred to here refers to an AI that calculates the score (confidence) of each feature (or its product) for the objective variable. Examples of explainable AI include frequent pattern mining and logistic regression.

[0115] Then, the interpretation rule generating unit 33 generates the interpretation rule 22A based on the list of the calculated scores of each feature amount. As an example, the interpretation rule generating unit 33 adds the specified relationship, feature amount, and score to the interpretation rule 22A in association with each other.

[0116] Fig. 19 is a diagram illustrating an example of generation of an interpretation rule according to the embodiment 5. As illustrated in Fig. 19, it is assumed that the interpretation rule generating unit 33 acquires a resource set list and "increase" as a relationship to be extracted (extraction relationship).

[0117] The interpretation rule generating unit 33 extracts, from the knowledge data 24, a feature table F corresponding to the resource set included in the resource set list and the extraction relationship "increase".

[0118] Then, the interpretation rule generation unit 33 uses the extracted feature table F as input, applies explainable AI with a variable representing the presence or absence of the extracted relationship "increase" as the objective variable, and calculates a list of scores for each feature included in the feature table F. Here, the score (confidence) for feature 1 is calculated to be "0.8", and the score (confidence) for feature 2 is calculated to be "0.6".

[0119] Then, the interpretation rule generation unit 33 generates the interpretation rule 22A based on the list of the calculated scores of each feature amount. Here, for the interpretation rule 22A, the interpretation rule generation unit 33 sets the extracted relationship "increase" as a relationship item, the feature 1 and the feature 2 as rule content items, and adds the scores (confidence) corresponding to the feature 1 and the feature 2 as score items to the interpretation rule 22A.

[0120] The interpretation rule generating unit 33 then performs the same process for the other extraction relations to generate the interpretation rule 22A.

[0121] Fig. 20 is a diagram showing an example of a flowchart of an interpretation rule generation process according to the fifth embodiment. As shown in Fig. 20, the interpretation rule generation unit 33 acquires a relationship r to be extracted and a resource list S_list (step S61). Then, the interpretation rule generation unit 33 extracts a feature amount table F for the resource set list from the resource set list S_list and the relationship r to be extracted, for example, by using the knowledge data 24 (step S62).

[0122] Then, the interpretation rule generation unit 33 applies the explainable AI to the feature table F using the variable indicating the presence or absence of the relationship r as the objective variable (step S63). Then, the interpretation rule generation unit 33 obtains a list score_list of the scores of each feature (step S64). Then, the interpretation rule generation unit 33 adds the list score_list of the scores of each feature to the interpretation rule 22A (step S65).

[0123] Then, the interpretation rule generating unit 33 judges whether or not all the relationships to be extracted have been acquired (step S66). If it is judged that all the relationships to be extracted have not been acquired (step S66; No), the interpretation rule generating unit 33 proceeds to step S61 to acquire the next relationship to be extracted.

[0124] On the other hand, if it is determined that all the relationships to be extracted have been acquired (step S66; Yes), the interpretation rule generating section 33 ends the interpretation rule generating process.

[0125] [Effects of Example 5] Thus, according to the fifth embodiment, the information processing device 1 further receives a list of resource pairs and the relationship of the extraction target. The information processing device 1 extracts features corresponding to the resource pairs included in the list and the relationship of the extraction target from the knowledge data 24. The information processing device 1 inputs the extracted features and applies an explainable AI with a variable indicating the presence or absence of the relationship of the extraction target as a target variable to output the certainty of the features. Then, the information processing device 1 adds information to the interpretation rule 22A, the information being information in which the extracted features are rules, the certainty of the output features is the likelihood of being existing knowledge, and information in which the relationship of the extraction target is associated. In this way, the information processing device 1 can generate the interpretation rule 22A by using the knowledge data 24. EXAMPLES

[0126] Incidentally, it has been described that the information processing device 1 according to the first embodiment interprets a hypothesis using the hypothesis interpretation DB 21 and the interpretation rule 22, and presents evidence and a score as an interpretation result. In the sixth embodiment, an information processing device 1 that presents an interpretation result will be described.

[0127] [Functional configuration of information processing device] Fig. 21 is a block diagram showing an example of a functional configuration of an information processing device according to Example 6. Note that, regarding the functional configuration of the information processing device 1 according to Example 6, the same components as those of the information processing device 1 shown in Fig. 18 are indicated with the same reference numerals, and a description of the overlapping configurations and operations will be omitted. Example 5 differs from Example 6 in that a hypothesis interpretation DB 41 and a hypothesis display unit 42 are added. Note that the hypothesis interpretation DB 41 is an example of a fourth storage unit.

[0128] The hypothesis interpretation DB 41 stores an interpretation result obtained by interpreting a hypothesis by the hypothesis interpretation unit 11. Here, an example of the hypothesis interpretation DB 41 will be described with reference to Fig. 22. Fig. 22 is a diagram showing an example of the hypothesis interpretation DB according to the sixth embodiment.

[0129] As shown in FIG. 22, the hypothesis interpretation DB 41 stores a hypothesis ID, a hypothesis structure, a hypothesis content, and an interpretation / evidence in association with each other. The hypothesis ID is an ID that uniquely identifies the hypothesis to be interpreted. The hypothesis structure corresponds to the relationship between the resources indicated by the hypothesis. The hypothesis content corresponds to multiple resources linked to the hypothesis structure. The interpretation / evidence is set with the rule ID and score of the rule used to interpret the hypothesis in the interpretation rules 22A, and the evidence linked to the rule.

[0130] As an example, when the hypothesis ID is “1”, the hypothesis structure is stored as “increase”, the hypothesis content is stored as “{“Gene1”:A, “Gene2”:B}”, and the interpretation / evidence is stored as “match score: 1 rule ID: 0, evidence: pmid789”.

[0131] The hypothesis display unit 42 displays the interpretation results for the hypothesis. For example, when the hypothesis display unit 42 receives a hypothesis ID specified by a user interface (UI), it extracts interpretations / evidence corresponding to the hypothesis ID from the hypothesis interpretation DB 41. Then, the hypothesis display unit 42 displays the extracted interpretations / evidence on the screen. Note that the hypothesis display unit 42 can filter and sort according to the interpretation results of the hypothesis using the UI.

[0132] FIG. 23 is a diagram showing an example of a screen image. The right diagram in FIG. 23 displays a menu showing one UI. The menu shows a filter list box and a sort list box. Here, "those with existing knowledge" is selected as the filter. "order by largest IF of existing knowledge" is selected as the sort. The hypothesis display unit 42 filters and extracts interpretation results of hypotheses from the hypothesis interpretation DB 41 according to the selected filter "those with existing knowledge". The hypothesis display unit 42 sorts the filtered and extracted interpretation results in the selected sort "order by largest IF of existing knowledge". The hypothesis display unit 42 then displays the filtered and sorted interpretation results on the screen.

[0133] The left diagram of Fig. 23 shows the interpretation results displayed by the hypothesis display unit 42. Here, the scores of the hypotheses are shown.

[0134] Fig. 24 is a diagram showing another example of a screen image. The left diagram in Fig. 24 shows a graph combining hypotheses. For example, nodes such as "EBF1" and "HLX" indicate gene IDs. The arrow and numbers between "EBF1" and "HLX" indicate the relationship between the genes. "EBF1 → HLX" is one hypothesis and has a unique hypothesis ID.

[0135] When, for example, an arrow portion of such a graph is clicked, the hypothesis display unit 42 receives the hypothesis ID corresponding to the clicked arrow portion. The hypothesis display unit 42 extracts the interpretation / evidence corresponding to the received hypothesis ID from the hypothesis interpretation DB 41. Then, the hypothesis display unit 42 displays the extracted interpretation / evidence on the screen. Here, for the hypothesis "EBF1→HLX", the interpretation / evidence "match score: 0.8 rule ID: 1, evidence: pmid789 rule ID: 2, evidence: pmid1234" is displayed.

[0136] [Effects of Example 6] In this way, according to the sixth embodiment, the information processing device 1 extracts and presents evidence information for a target hypothesis and the likelihood that the information is existing knowledge from the hypothesis interpretation DB 41 based on the user interface. This allows the information processing device 1 to easily present the interpretation result of the hypothesis to be interpreted. As a result, the user can efficiently select the hypothesis to be interpreted.

[0137] Note that each component of the illustrated information processing device 1 does not necessarily have to be physically configured as illustrated. In other words, the specific form of distribution and integration of the information processing device 1 is not limited to that shown in the figure, and all or a part of it can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc. For example, the hypothesis interpretation unit 11 may be distributed into a functional unit that interprets hypotheses, a functional unit that collects interpretation results, and a functional unit that presents them. In addition, a storage unit (not shown) that stores the hypothesis interpretation DB 21 and the interpretation rules 22 may be connected as an external device of the information processing device 1 via a network.

[0138] Moreover, the various processes described in the above embodiments can be realized by executing a prepared program on a computer such as a personal computer or a workstation. Therefore, an example of a computer that executes an information processing program that realizes the same functions as the information processing device 1 shown in Fig. 1 will be described below. Here, an information processing program that realizes the same functions as the information processing device 1 will be described as an example. Fig. 25 is a diagram showing an example of a computer that executes an information processing program.

[0139] 25, the computer 200 has a CPU 203 that executes various arithmetic processes, an input device 215 that accepts data input from a user, and a display control unit 207 that controls a display device 209. The computer 200 also has a drive device 213 that reads programs and the like from a storage medium, and a communication control unit 217 that transmits and receives data to and from other computers via a network. The computer 200 also has a memory 201 that temporarily stores various information, and an HDD (Hard Disk Drive) 205. The memory 201, the CPU 203, the HDD 205, the display control unit 207, the drive device 213, the input device 215, and the communication control unit 217 are connected by a bus 219.

[0140] The drive device 213 is, for example, a device for the removable disk 210. The HDD 205 stores an information processing program 205a and information processing related information 205b.

[0141] The CPU 203 reads out the information processing program 205a, expands it in the memory 201, and executes it as a process. Such a process corresponds to each functional unit of the information processing device 1. The information processing related information 205b corresponds to, for example, the hypothesis interpretation DB 21, the interpretation rule 22, and the query information 23. And, for example, the removable disk 210 stores each piece of information such as the information processing program 205a.

[0142] It should be noted that the information processing program 205a does not necessarily have to be stored in the HDD 205 from the beginning. For example, the program may be stored in a "portable physical medium" such as a flexible disk (FD), a CD-ROM (Compact Disk Read Only Memory), a DVD (Digital Versatile Disk), a magneto-optical disk, or an IC (Integrated Circuit) card that is inserted into the computer 200. Then, the computer 200 may read out the information processing program 205a from these and execute it. [Explanation of symbols]

[0143] 1. Information processing device 11,11A Hypothesis Interpretation Section 12 DB generation section 21 Hypothesis interpretation database 21a Knowledge Table 21b Evidence Table 22, 22A Interpretation Rules 23 Query Information 24 Knowledge Data 31 Relationship Extraction Unit 32 DB Integration Department 33 Interpretation rule generation unit 41 Hypothesis Interpretation DB 42 Hypothesis display section

Claims

1. Receive the hypothesis to be interpreted, acquiring, for each piece of knowledge indicating a plurality of resources and a relationship between the resources, basis information that is the basis of the knowledge and a rule identifier that is linked to a rule used to interpret a hypothesis, basis information and a rule identifier that correspond to the hypothesis to be interpreted, using a first storage unit that includes the basis information of the knowledge and a rule identifier that is linked to a rule used to interpret the hypothesis; A second storage unit that stores the likelihood that the rule and the hypothesis are existing knowledge for each rule identifier is used to obtain the likelihood that the rule and the hypothesis are existing knowledge corresponding to the obtained rule identifier. An information processing program that causes a computer to execute processing.

2. Present the obtained grounds information and the likelihood that the hypothesis is existing knowledge for the interpretation target.

2. The information processing program according to claim 1,

3. It also receives a set of resources, acquires, for the set of resources, basis information from a third storage unit in which basis information that is a basis for knowledge is associated with each piece of knowledge including a subject, an object, and a predicate that indicates a relationship between the subject and the object, based on the rule included in the second storage unit; For the set of resources, information including the relationship between the resources obtained from the rule in the second storage unit, the basis information, and a rule identifier associated with the rule is added to the first storage unit.

2. The information processing program according to claim 1,

4. generating the third storage unit using a predetermined relationship extraction process from natural language and using the target document as evidence information; A predetermined integration process for integrating a plurality of existing databases is used to integrate the plurality of target databases, and the third storage unit is generated using the integrated database as evidence information.

4. The information processing program according to claim 3.

5. It also receives a list of resource tuples and the relationships to be extracted; extracting features corresponding to the resource pairs included in the list and the relationships to be extracted from a third storage unit that links basis information that is the basis of knowledge for each piece of knowledge including a subject, an object, and a predicate that indicates a relationship between the subject and the object; The extracted features are input, and an explainable AI is applied to which a variable indicating the presence or absence of a relationship between the extracted features is set as a target variable, to output a confidence level of the features. Add information in which the extracted feature quantity is the rule, the certainty of the output feature quantity is the likelihood that the feature quantity is the existing knowledge, and information in which the relationship of the extracted object is associated with the rule is added to the second storage unit.

2. The information processing program according to claim 1,

6. Furthermore, the acquired grounds information and the likelihood that the hypothesis to be interpreted is existing knowledge are stored in a fourth storage unit; extracting, from the fourth storage unit, the evidence information for the target hypothesis and the likelihood that the information is existing knowledge, and presenting the extracted evidence information and existing knowledge based on a user interface; 2. The information processing program according to claim 1,

7. Receive the hypothesis to be interpreted, using rules related to the hypothesis to be interpreted, which are included in a second storage unit including rules used to interpret a hypothesis and the likelihood that the hypothesis is existing knowledge for each rule identifier, to determine whether or not knowledge corresponding to the hypothesis to be interpreted exists in a third storage unit in which basis information that is the basis of knowledge is linked to each piece of knowledge including a subject, an object, and a predicate that indicates a relationship between the subject and the object; When it is determined that knowledge corresponding to the hypothesis to be interpreted exists in the third storage unit, grounds information corresponding to the knowledge and a likelihood that the knowledge is existing knowledge corresponding to a rule related to the hypothesis to be interpreted are acquired. An information processing program that causes a computer to execute processing.

8. a first acquisition unit that receives a hypothesis to be interpreted, and acquires, for each piece of knowledge indicating a plurality of resources and a relationship between the resources, basis information that is a basis for the knowledge and a rule identifier that is linked to a rule used to interpret the hypothesis, and the basis information and the rule identifier corresponding to the hypothesis to be interpreted; a second acquisition unit that acquires, for each rule identifier, a likelihood that the rule and the hypothesis are existing knowledge, using a second storage unit that includes a likelihood that the rule and the hypothesis are existing knowledge; 13. An information processing device comprising:

9. Receive the hypothesis to be interpreted, acquiring, for each piece of knowledge indicating a plurality of resources and a relationship between the resources, basis information that is the basis of the knowledge and a rule identifier that is linked to a rule used to interpret a hypothesis, basis information and a rule identifier that correspond to the hypothesis to be interpreted, using a first storage unit that includes the basis information of the knowledge and a rule identifier that is linked to a rule used to interpret the hypothesis; A second storage unit that stores the likelihood that the rule and the hypothesis are existing knowledge for each rule identifier is used to obtain the likelihood that the rule and the hypothesis are existing knowledge corresponding to the obtained rule identifier. An information processing method in which processing is performed by a computer.

Images