Historical culture knowledge fusion model construction method and related equipment

Abstract

The invention relates to the technical field of artificial intelligence, and provides a construction method of a historical culture knowledge fusion model and related equipment. The construction method of the historical and cultural knowledge fusion model comprises the steps of obtaining knowledge information of target historical and cultural domain data after knowledge extraction, and determining theme concept information of historical and cultural knowledge fusion, the target historical and cultural field data is historical and cultural field data of a plurality of data sources normalized by a preset multi-modal model, and the topic concept information comprises a topic field and a concept type; performing entity alignment based on the theme concept information and the knowledge information to obtain aligned knowledge information; performing multi-level fusion on the aligned knowledge information to obtain fused knowledge information; and auditing the fused knowledge information, and constructing a target historical and cultural knowledge fusion model based on the fused knowledge information passing the auditing. According to the method, the knowledge fusion accuracy of the historical and cultural knowledge fusion model is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method for constructing a historical and cultural knowledge fusion model and related equipment. Background Art

[0002] In order to facilitate the management and viewing of historical and cultural data, a corresponding knowledge model is constructed to integrate historical and cultural data. At present, historical and cultural data are generally integrated together to construct a corresponding knowledge fusion model. However, due to the characteristics of historical and cultural data, the above method leads to a low knowledge fusion accuracy of the constructed historical and cultural knowledge fusion model. Summary of the invention

[0003] The present invention provides a method for constructing a historical and cultural knowledge fusion model and related equipment, which improves the knowledge fusion accuracy of the historical and cultural knowledge fusion model by fusing historical and cultural data from multiple data sources from multiple angles in the historical and cultural field and conducting audits.

[0004] The first aspect of the present invention provides a method for constructing a historical and cultural knowledge fusion model, comprising:

[0005] Acquire knowledge information of target historical and cultural field data after knowledge extraction, and determine the subject concept information of historical and cultural knowledge fusion, wherein the target historical and cultural field data is historical and cultural field data from multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes subject field and concept type;

[0006] Perform entity alignment based on the subject concept information and the knowledge information to obtain aligned knowledge information;

[0007] Performing multi-level fusion on the aligned knowledge information to obtain fused knowledge information;

[0008] The integrated knowledge information is reviewed, and a target historical and cultural knowledge fusion model is constructed based on the reviewed integrated knowledge information.

[0009] A second aspect of the present invention provides a device for constructing a historical and cultural knowledge fusion model, comprising:

[0010] A determination module is used to obtain knowledge information of the target historical and cultural field data after knowledge extraction, and to determine the subject concept information of the historical and cultural knowledge fusion, wherein the target historical and cultural field data is the historical and cultural field data of multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes a subject field and a concept type;

[0011] An alignment module, used for performing entity alignment based on the subject concept information and the knowledge information to obtain aligned knowledge information;

[0012] A fusion module, used for performing multi-level fusion on the aligned knowledge information to obtain fused knowledge information;

[0013] The review and construction module is used to review the integrated knowledge information and construct a target historical and cultural knowledge fusion model based on the integrated knowledge information that has passed the review.

[0014] The third aspect of the present invention provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores a computer program; the at least one processor calls the computer program in the memory so that the electronic device executes the above-mentioned method for constructing a historical and cultural knowledge fusion model.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium, in which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer executes the method for constructing the historical and cultural knowledge fusion model described above.

[0016] In the technical solution provided by the present invention, the knowledge information of the target historical and cultural field data after knowledge extraction is obtained, and the subject concept information of the historical and cultural knowledge fusion is determined, wherein the target historical and cultural field data is the historical and cultural field data of multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes the subject field and the concept type; entity alignment is performed based on the subject concept information and the knowledge information to obtain the aligned knowledge information; multi-level fusion is performed on the aligned knowledge information to obtain the fused knowledge information; the fused knowledge information is reviewed, and the target historical and cultural knowledge fusion model is constructed based on the reviewed fused knowledge information. In the embodiment of the present invention, the knowledge fusion accuracy of the historical and cultural knowledge fusion model is improved by fusing the historical and cultural field data of multiple data sources from multiple angles in the historical and cultural field and reviewing them. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of an embodiment of a method for constructing a historical and cultural knowledge fusion model in an embodiment of the present invention;

[0018] Figure 2 It is another schematic diagram of an embodiment of the method for constructing a historical and cultural knowledge fusion model in an embodiment of the present invention;

[0019] Figure 3 A schematic diagram of an embodiment of a device for constructing a historical and cultural knowledge fusion model in an embodiment of the present invention;

[0020] Figure 4 FIG. 4 is a schematic diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The embodiment of the present invention provides a method for constructing a historical and cultural knowledge fusion model and related equipment, which improves the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0022] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] It is understandable that the execution subject of the present invention may be a device for constructing a historical and cultural knowledge fusion model, or may be a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0024] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the method for constructing a historical and cultural knowledge fusion model in an embodiment of the present invention includes:

[0025] 101. Obtaining knowledge information of the target historical and cultural field data after knowledge extraction, and determining the subject concept information of the historical and cultural knowledge fusion, wherein the target historical and cultural field data is historical and cultural field data from multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes subject fields and concept types;

[0026] Among them, as an example but not limitation, the above-mentioned multiple data sources include but are not limited to institutional collection data (for example, literature, manuscripts, pictures, etc. collected by institutions such as libraries, museums and archives), digital historical and cultural resources publicly available on the Internet (for example, digital historical and cultural resources such as digital libraries and digital museums), field investigation data (for example, the results of field investigations and excavations of historical and cultural sites and ancient monuments), oral historical and cultural data (for example, historical and cultural materials obtained through interviews with oral stories and traditional culture of the older generation) and academic research historical and cultural data (for example, academic research results and reports published by scholars, experts, research institutions, etc.); the historical and cultural field data of the above-mentioned multiple data sources correspond to a variety of data formats, such as images, videos, texts and voices.

[0027] As an example but not limitation, the above-mentioned preset multimodal model is a model for processing data in multiple data formats. For example, the above-mentioned preset multimodal model includes an image processing sub-model, a video processing sub-model, a text processing sub-model and a speech processing sub-model. The video processing sub-model, the image processing sub-model and the text processing sub-model are connected in sequence, that is, the output of the video processing sub-model is the input of the image processing sub-model, and the output of the image processing sub-model is the input of the text processing sub-model, that is, the image processing sub-model verifies or performs text conversion on the recognition result of the video processing sub-model, and the text processing sub-model verifies or performs text conversion on the recognition result of the image processing sub-model, and the speech processing sub-model is connected to the text processing sub-model, and the output of the speech processing sub-model is the text processing sub-model, that is, the text processing sub-model verifies or performs text conversion on the recognition result of the speech processing sub-model.

[0028] As an example but not limitation, the above-mentioned knowledge information includes but is not limited to entities, entity relationships and entity attributes; the above-mentioned subject areas can be understood as subject areas in the field of history, such as ancient history, modern history, cultural history and economic history; the above-mentioned concept types can be understood as the classification method of specific concepts in the subject areas in the field of history.

[0029] 102. Perform entity alignment based on subject concept information and knowledge information to obtain aligned knowledge information;

[0030] As an example but not limitation, the corresponding target rules are matched from the preset rule library through the subject concept information, wherein the target rules are items corresponding to each entity type that can be aligned or matched, and the target rules include alignment target items corresponding to each entity type, for example, the name, alias, title and characters of historical figures, and the alignment target item refers to the preset target entity item for entity alignment, for example, each entity corresponding to the historical figure is aligned to the entity corresponding to the name; the corresponding rules are matched from the target rules according to the entity type of each entity in the knowledge information; the corresponding entities are aligned to the entities corresponding to the alignment target items in the matched rules according to the matched rules, to obtain the aligned knowledge information.

[0031] As an example but not limitation, similar entities in different data sources are obtained; similar entities in different data sources are aligned to the same preset entity to obtain aligned knowledge information. Specifically, knowledge information is matched based on a preset entity matching algorithm (e.g., a string matching algorithm, a semantic matching algorithm, and a structured matching algorithm) to obtain initial entities from multiple data sources; initial entities from multiple data sources are adjusted based on a preset similarity algorithm (e.g., a text similarity algorithm, a semantic similarity algorithm, a contextual similarity algorithm, and a co-occurrence frequency similarity algorithm) to obtain aligned knowledge information.

[0032] By aligning entities based on subject concept information and knowledge information, the problems of information redundancy and overlap caused by the diversity and heterogeneity of multiple data sources are solved.

[0033] 103. Perform multi-level fusion on the aligned knowledge information to obtain fused knowledge information;

[0034] As an example but not limitation, a multi-level fusion strategy is determined based on the subject concept information, wherein the multi-level fusion strategy includes the content and method corresponding to each level of fusion; the aligned knowledge information is fused based on the multi-level fusion strategy to obtain the fused knowledge information. Specifically, the aligned knowledge information is attribute fused and relationship fused based on the multi-level fusion strategy to obtain the fused knowledge; the fused knowledge is inferred based on the historical domain rules and historical domain logic in the multi-level fusion strategy to fill the fused knowledge and obtain the fused knowledge information. Among them, the above-mentioned relationship fusion can be to merge and unify the attributes of the same entity (i.e., the entity in the aligned knowledge information) in multiple data sources, for example, to integrate the attributes such as date of birth and nationality in different data sources of the same historical figure; the above-mentioned relationship fusion can be understood as integrating the same relationships in the aligned knowledge information to enrich and improve the relationship representation of the knowledge graph, for example, to merge the relevant relationships between different historical events to form a more comprehensive historical event relationship network.

[0035] As an example but not a limitation, the aforementioned integrated knowledge information may be a knowledge graph.

[0036] 104. Review the integrated knowledge information, and build a target historical and cultural knowledge fusion model based on the reviewed integrated knowledge information.

[0037] As an example but not limitation, a preset historical and cultural knowledge review model is called to review and score the integrated knowledge information to obtain an review score, wherein the historical and cultural knowledge review model is a model based on an artificial intelligence neural network for reviewing the consistency and logic between entities, attributes and relationships of historical and cultural knowledge; if the review score is greater than the preset score, the review of the integrated knowledge information is determined to have passed, and a historical and cultural knowledge fusion model is constructed based on the integrated knowledge information; if the review score is less than or equal to the preset score, the integrated knowledge information is sent to the review end, and the reviewed integrated knowledge information returned by the review end is received to obtain the reviewed integrated knowledge information, wherein the reviewed integrated knowledge information returned by the review end is a hierarchical review of the integrated knowledge information by the auditors at the review end according to the review level, and the integrated knowledge information that fails the review is trimmed, and the final review level corresponds to the trimmed integrated knowledge information.

[0038] In one implementation, after the above step 104, the knowledge information to be processed is obtained after knowledge extraction of the historical and cultural field data to be processed, and the target subject concept information of the historical and cultural knowledge fusion is determined, wherein the historical and cultural field data to be processed are historical and cultural field data of multiple data sources to be processed that have been normalized by a preset multimodal model, and the target subject concept information includes a target subject field and a target concept type; the target historical and cultural knowledge fusion model is called to construct a target knowledge graph based on the knowledge information to be processed and the target subject concept information; the historical and cultural field data corresponding to each entity of the target knowledge graph is obtained from the historical and cultural field data to be processed; the historical and cultural field data corresponding to each entity is inserted or linked to the corresponding position of each entity in the target knowledge graph in the form of a hover prompt according to a preset display method.

[0039] By constructing the target knowledge graph through the target historical and cultural knowledge fusion model, the quality, consistency and richness of the historical and cultural knowledge in the target knowledge graph are improved.

[0040] In one implementation, after the above-mentioned step of inserting or linking the historical and cultural field data corresponding to each entity to the position corresponding to each entity in the target knowledge graph in the form of a hover prompt, it also includes: in response to a hover operation of the target knowledge graph, displaying the historical and cultural field data corresponding to the hover operation position in a corresponding data format through a target display method, wherein the target display method includes but is not limited to a floating display method and a pop-up display method.

[0041] By presenting historical and cultural data through hover display operations and multiple data formats, on the basis of constructing a high-quality target knowledge graph based on the target historical and cultural knowledge fusion model, the data presentation method and data presentation angle of the target knowledge graph are enriched, and the convenience and interpretability are improved.

[0042] In the embodiment of the present invention, the historical and cultural data from multiple data sources are integrated from multiple angles in the historical and cultural field and then reviewed, thereby improving the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0043] See also Figure 2 Another embodiment of the method for constructing the historical and cultural knowledge fusion model in the embodiment of the present invention includes:

[0044] 201. Obtaining knowledge information of target historical and cultural field data after knowledge extraction, and determining subject concept information of historical and cultural knowledge fusion, wherein the target historical and cultural field data is historical and cultural field data from multiple data sources that has been normalized by a preset multimodal model, and the subject concept information includes subject field and concept type;

[0045] In one implementation, the above step 201 includes: obtaining historical and cultural field data from multiple data sources, and the data formats of the historical and cultural field data from multiple data sources; calling a preset multimodal model to standardize the historical and cultural field data from multiple data sources based on the data format to obtain standardized historical and cultural field data; performing data preprocessing on the standardized historical and cultural field data to obtain preprocessed historical and cultural field data; performing knowledge extraction on the preprocessed historical and cultural field data to obtain knowledge information; and determining thematic concept information of the fusion of historical and cultural knowledge.

[0046] As an example but not limitation, after obtaining historical and cultural data from multiple data sources, data cleaning and security testing are performed on the historical and cultural data obtained from multiple data sources to perform data preprocessing on the historical and cultural data after standardization, so as to ensure the quality and security of the historical and cultural data from multiple data sources, thereby helping to improve the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0047] Among them, the above-mentioned calling of the preset multimodal model and standardizing the historical and cultural data from multiple data sources based on the data format can be understood as converting the historical and cultural data in different data formats from multiple data sources into the same data format through the preset multimodal model.

[0048] Among them, as an example but not limitation, the above-mentioned step of extracting knowledge from the preprocessed historical and cultural data to obtain knowledge information includes: calling a preset language model and machine learning algorithm, based on the dictionary and rules corresponding to the field of history, performing entity recognition and entity extraction on the preprocessed historical and cultural data, to obtain initial entities from multiple data sources, wherein the language model is used to perform context-based semantic recognition on the preprocessed historical and cultural data, and the machine learning algorithm is used to classify or cluster the information recognized by the language model; adjusting the initial entities of multiple data sources based on thematic concept information, and annotating key entities in the preprocessed historical and cultural data to obtain target entities from multiple data sources, wherein each target entity is used as a node in the knowledge graph; through a preset natural language processing algorithm, based on a preset historical The dictionary and knowledge base corresponding to the field of history are used to extract relationships from the pre-processed historical and cultural data to obtain relationship information from multiple data sources, wherein a natural language processing algorithm is used to perform lexical analysis, syntactic analysis, semantic role labeling, polysemous word processing and context understanding on the pre-processed historical and cultural data; through the natural language processing algorithm, the sentence structure and grammatical rules of the pre-processed historical and cultural data are parsed based on the subject concept information to locate the location of the attribute value and obtain the attribute location of multiple data sources; based on the dictionary and knowledge base corresponding to the field of history, attributes of the pre-processed historical and cultural data are extracted and classified based on the attribute location to obtain the attribute information of multiple data sources; the target entities, relationship information and attribute information of multiple data sources are determined as the knowledge information of the target historical and cultural data after knowledge extraction.

[0049] By preprocessing, multimodal processing and knowledge extraction of historical and cultural data from multiple data sources, and determining the subject concept information of historical and cultural knowledge fusion, the quality of knowledge information is guaranteed, which helps to improve the accuracy of subsequent historical and cultural knowledge fusion.

[0050] 202. Classify knowledge information based on subject concept information to obtain multiple types of information;

[0051] As an example but not limitation, a preset classifier is called to classify entities in knowledge information based on subject concept information to obtain entity information corresponding to multiple types, i.e., multi-category information, wherein the entity information corresponding to each type includes the entity, the relationship and attributes corresponding to the entity, and the types include but are not limited to people, events, places, eras and political systems.

[0052] 203. Determine a target individual corresponding to each type of information in the multiple types of information based on a preset genetic algorithm, wherein the target individual indicates an optimal entity alignment solution;

[0053] As an example but not limitation, an initial population is randomly generated based on each type of information in multiple types of information to obtain an initial population corresponding to each type of information, wherein each individual in the initial population indicates an entity alignment scheme; the initial population corresponding to each type of information is subjected to crossover mutation and proportional selection processing to obtain excellent individuals corresponding to each type of information; the excellent individuals corresponding to each type of information are subjected to stable iterative processing until the deviation values ​​of the excellent individuals obtained by the stable iterative processing converge, thereby obtaining target individuals corresponding to each type of information.

[0054] 204. Based on the target individuals corresponding to each type of information, the corresponding entities are integrated to obtain aligned knowledge information;

[0055] As an example but not limitation, the above step 204 specifically includes: based on the target individuals corresponding to each type of information, similar entities in the knowledge information are obtained, wherein similar entities can be understood as entities whose similarity is greater than a preset similarity threshold. Specifically, the similarity of entities between multiple data sources of the same type in the knowledge information can be calculated based on the similarity calculation factor in the target individuals corresponding to each type of information to obtain entity similarity, and entities whose entity similarity is greater than the preset similarity threshold are determined as similar entities corresponding to each type of information. The similarity calculation factor can be understood as a factor used to calculate similarity, such as data structure, semantics, context and co-occurrence frequency; according to the target individuals corresponding to each type of information, similar entities corresponding to each type of information are aligned to the preset standard entities corresponding to each type of information to obtain aligned knowledge information.

[0056] By aligning knowledge information based on subject concept information and genetic algorithms, the accuracy of entity matching and entity alignment is improved, and the problems of information redundancy and overlap caused by the diversity and heterogeneity of multiple data sources are further effectively solved.

[0057] 205. Perform attribute fusion on the aligned knowledge information to obtain initial fused knowledge information;

[0058] In one implementation, the above step 205 includes: dividing the attributes in the aligned knowledge information according to the aligned entities to obtain the attributes of each similar entity; classifying the attributes of each similar entity to obtain various attributes of each similar entity; integrating various attributes of each similar entity based on a preset historical and cultural field attribute strategy to obtain initial integrated knowledge information.

[0059] As an example but not limitation, the attributes of the above-mentioned similar entities can be understood as the attributes corresponding to the entities after the entity types are aligned; the above-mentioned preset historical and cultural field attribute strategy can be understood as the attribute fusion strategy corresponding to the subject concept information in the preset historical and cultural field, for example, the merging rules for the time attributes representing historical dynasties.

[0060] As an example but not a limitation, the above-mentioned attributes of various similar entities are merged, deduplicated and standardized based on the preset historical and cultural domain attribute strategy to obtain initial fused knowledge information. Specifically, the attention weight of each attribute in various attributes of various similar entities can be calculated based on the preset historical and cultural domain attribute strategy and the preset attention mechanism; the various attributes of various similar entities are weightedly fused according to the calculated attention mechanism to obtain the fused attributes; the corresponding feature representation (feature vector) is constructed based on the fused attributes to obtain the initial fused knowledge information.

[0061] By classifying the attributes of the aligned knowledge information and fusing the attributes based on the preset historical and cultural field attribute strategy, the accuracy and adaptability of the feature representation in the initial fused knowledge information are improved, as well as the completeness and accuracy of the initial fused knowledge information.

[0062] 206. Perform relation fusion on the initial fused knowledge information to obtain re-fused knowledge information;

[0063] In one implementation, the above step 206 includes: classifying the entity relationships in the initial fused knowledge information to obtain classified entity relationships; transforming the classified entity relationships based on a preset historical and cultural field relationship template to obtain integrated entity relationships; connecting the initial fused knowledge information based on a preset historical and cultural field relationship path and the integrated entity relationships to obtain a historical and cultural relationship network; optimizing the historical and cultural relationship network based on a preset optimization algorithm to obtain re-fused knowledge information.

[0064] As an example but not limitation, based on a preset relationship library in the field of history, the entity relationships in the initial fused knowledge information are matched and classified to obtain the classified entity relationships; the above-mentioned historical and cultural field relationship template is a template corresponding to the merging and unification of similar relationships, for example, the similar relationship is that A and B are husband and wife, and A is B's husband, then the corresponding historical and cultural field relationship template is A is B's husband; the preset historical and cultural field relationship path is the relationship path between entities constructed by experts in various historical fields based on personal experience knowledge and public and authentic historical and cultural knowledge, for example, the relationship between historical dynasties and the social class structure and ritual system corresponding to the historical dynasties; the above-mentioned optimization algorithm can be a local optimal solution algorithm, a particle swarm optimization algorithm and an ant colony algorithm, which are not specifically limited here, and the historical and cultural relationship network is verified and screened by the preset optimization algorithm and thematic concept information to determine the optimal historical and cultural relationship network.

[0065] By classifying the initial fused knowledge information, transforming the relationship templates and optimizing the relationship network, it is possible to solve the problem in the field of history that only the data corresponding to the entities, relationships and attributes of people, places and events with relatively high recognition can be integrated, resulting in low knowledge fusion accuracy. It is possible to perform relationship integration from multiple angles and in an optimal solution manner, thereby improving the accuracy of knowledge fusion and thus improving the richness and accuracy of the re-integrated knowledge information.

[0066] 207. Based on the preset historical and cultural field rules, the re-integrated knowledge information is integrated to obtain the integrated knowledge information;

[0067] By fusing the aligned knowledge information with attributes, relationships, and rules based on the historical and cultural fields, multi-angle fusion is achieved. This can largely solve the problem of only being able to fuse historical and cultural knowledge data with relatively high recognition, improve the accuracy and richness of the integrated knowledge information, and thus help improve the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0068] In one implementation, the above step 207 includes: traversing the re-integrated knowledge information based on preset historical and cultural field rules to obtain traversal results; adjusting and filling the re-integrated knowledge information through the traversal results to obtain integrated knowledge information.

[0069] As an example but not limitation, based on the rules of the historical and cultural field, the re-integrated knowledge information is inferred to obtain reasoning information, wherein the above-mentioned historical and cultural field rules are the experiential knowledge and rules of experts in the field of history; through the preset historical and cultural field rules and reasoning information, the re-integrated knowledge information is traversed to proofread, adjust and fill the traversal of the re-integrated knowledge information to eliminate conflicting entities and relationships, as well as to infer and supplement the missing historical and cultural knowledge to obtain the integrated knowledge information.

[0070] The re-integrated knowledge information is proofread, adjusted and supplemented through the rules of historical and cultural fields, conflicting entities and relationships are eliminated, and missing historical and cultural knowledge is inferred and supplemented. The re-integrated knowledge information is fused, integrated and supplemented from different data sources, which further effectively improves the accuracy of the re-integrated knowledge information, which in turn helps to improve the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0071] 208. Review the integrated knowledge information and build a target historical and cultural knowledge fusion model based on the reviewed integrated knowledge information.

[0072] In one implementation, the above step 208 includes: performing similarity calculation based on a preset historical knowledge graph library and the integrated knowledge information to obtain the similarity of each entity sequence in the integrated knowledge information; determining the target entity sequence through a preset threshold and the similarity of each entity sequence, wherein the target entity sequence is an entity sequence whose similarity is less than or equal to the preset threshold; performing a pre-review on the target entity sequence; if the target entity sequence passes the pre-review, then constructing a target historical and cultural knowledge fusion model based on the integrated knowledge information.

[0073] As an example but not limitation, if the preset historical knowledge graph library contains multiple entity sequences of similar type to the entity sequences in the integrated knowledge information, the similarity between each entity sequence in the integrated knowledge information and the multiple entity sequences of similar type in the historical knowledge graph library is calculated to obtain multiple calculated similarities; based on the preset priority, the calculated multiple similarities are weighted summed or the calculated multiple similarities are calculated as the arithmetic mean to obtain the similarity of each entity sequence in the integrated knowledge information; or, if the preset historical knowledge graph library contains multiple entity sequences of similar type to the entity sequences in the integrated knowledge information, the similarity between each entity sequence in the integrated knowledge information and the multiple entity sequences of similar type in the historical knowledge graph library is calculated to obtain multiple calculated similarities; the largest similarity among the calculated multiple similarities is used as the similarity of each entity sequence in the integrated knowledge information.

[0074] As an example but not limitation, the content of the pre-review of the target entity sequence includes but is not limited to: potential errors, conflicts or inconsistencies, consistency and logic between entities, attributes and relationships; taking the entity sequence in the fused knowledge information whose similarity is less than or equal to a preset threshold as the target entity sequence; if the target entity sequence passes the pre-review, then constructing an initial historical and cultural knowledge fusion model with supervised learning function based on the fused knowledge information; calculating the loss value of the initial historical and cultural knowledge fusion model; judging whether the loss value is greater than the preset loss value; if the loss value is greater than the preset loss value, then taking the initial historical and cultural knowledge fusion model as the target historical and cultural knowledge fusion model; if the loss value is less than or equal to the preset loss value, then training the initial historical and cultural knowledge fusion model until the loss value is greater than the preset loss value to obtain the target historical and cultural knowledge fusion model.

[0075] By reviewing the entity sequences in the fused knowledge information whose similarity is less than or equal to the preset threshold, the computational complexity of the review can be reduced, the processing efficiency can be improved, the reliability of the data for constructing the target historical and cultural knowledge fusion model can be improved, and the knowledge fusion accuracy of the target historical and cultural knowledge fusion model can also be improved.

[0076] In one implementation, after the above-mentioned step of pre-reviewing the target entity sequence, it also includes: if the target entity sequence includes an entity sequence that fails the pre-review, trimming the entity sequence that fails the pre-review to obtain a trimmed entity sequence; reviewing the trimmed entity sequence according to a preset number of reviews to obtain a target trimmed entity sequence; and constructing a target historical and cultural knowledge fusion model based on the target trimmed entity sequence, the entity sequence that passes the pre-review, and the entity sequence whose similarity is greater than a preset threshold.

[0077] As an example but not limitation, the above-mentioned entity sequence that fails the pre-audit can be understood as an entity sequence with potential errors, conflicts or inconsistencies, as well as an entity sequence with problems in consistency and logic between entities, attributes and relationships; in response to a trimming operation, trimming information (for example, manually input trimming content) is obtained, and the position to be trimmed in the entity sequence that fails the pre-audit is determined, and the original information in the position to be trimmed in the entity sequence that fails the pre-audit is replaced with the trimming information to implement the audit of the trimmed entity sequence to obtain a target trimmed entity sequence; or, the audit information corresponding to the entity sequence that fails the pre-audit is obtained, and the corresponding trimming content is matched from a preset database based on the audit information; the entity sequence that fails the pre-audit is trimmed based on the trimming content to obtain a trimmed entity sequence; according to a preset number of audits, the trimmed entity sequence is audited at multiple levels based on preset audit levels to obtain a target trimmed entity sequence to ensure that errors are completely eliminated and high-quality knowledge is obtained.

[0078] By trimming the entity sequence that fails the pre-review, and performing multiple rounds of review and error correction iterations on the trimmed entity sequence according to a preset number of reviews, the computational complexity of the review can be reduced, processing efficiency can be improved, the quality and credibility of the data for constructing the target historical and cultural knowledge fusion model can be improved, and misleading and erroneous knowledge can be reduced, further effectively improving the knowledge fusion accuracy of the target historical and cultural knowledge fusion model.

[0079] In the embodiment of the present invention, the historical and cultural data from multiple data sources are integrated from multiple angles in the historical and cultural field and then reviewed, thereby improving the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0080] The above describes the method for constructing the historical and cultural knowledge fusion model in the embodiment of the present invention. The following describes the device for constructing the historical and cultural knowledge fusion model in the embodiment of the present invention. Figure 3 In one embodiment of the present invention, a device for constructing a historical and cultural knowledge fusion model includes:

[0081] The determination module 301 is used to obtain knowledge information of the target historical and cultural field data after knowledge extraction, and to determine the subject concept information of the historical and cultural knowledge fusion, wherein the target historical and cultural field data is the historical and cultural field data of multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes the subject field and the concept type;

[0082] An alignment module 302 is used to perform entity alignment based on subject concept information and knowledge information to obtain aligned knowledge information;

[0083] A fusion module 303 is used to perform multi-level fusion on the aligned knowledge information to obtain fused knowledge information;

[0084] The review and construction module 304 is used to review the integrated knowledge information and construct a target historical and cultural knowledge fusion model based on the reviewed integrated knowledge information.

[0085] In the embodiment of the present invention, the historical and cultural data from multiple data sources are integrated from multiple angles in the historical and cultural field and then reviewed, thereby improving the knowledge fusion accuracy of the historical and cultural knowledge fusion model.

[0086] Optionally, the fusion module 303 may also be specifically used for:

[0087] Perform attribute fusion on the aligned knowledge information to obtain initial fused knowledge information;

[0088] Perform relational fusion on the initial fused knowledge information to obtain re-fused knowledge information;

[0089] Based on the preset historical and cultural field rules, the knowledge information is integrated again to obtain the integrated knowledge information.

[0090] Optionally, the fusion module 303 may also be specifically used for:

[0091] Divide the attributes in the aligned knowledge information according to the aligned entities to obtain the attributes of each similar entity;

[0092] Classify the attributes of each similar entity to obtain various attributes of each similar entity;

[0093] Based on the preset historical and cultural attribute strategy, various attributes of similar entities are integrated to obtain initial fusion knowledge information.

[0094] Optionally, the fusion module 303 may also be specifically used for:

[0095] Classify the entity relationships in the initial fused knowledge information to obtain the classified entity relationships;

[0096] Based on the preset historical and cultural field relationship template, the classified entity relationships are transformed to obtain the integrated entity relationships;

[0097] Based on the preset historical and cultural field relationship paths and the integrated entity relationships, the initial fusion knowledge information is connected to obtain a historical and cultural relationship network;

[0098] Based on the preset optimization algorithm, the historical and cultural relationship network is optimized to obtain the re-integrated knowledge information.

[0099] Optionally, the alignment module 302 may also be specifically configured to:

[0100] Classify knowledge information based on subject concept information to obtain multiple categories of information;

[0101] Determine a target individual corresponding to each type of information in multiple types of information based on a preset genetic algorithm, wherein the target individual indicates an optimal entity alignment solution;

[0102] Based on the target individuals corresponding to each type of information, the corresponding entities are integrated to obtain aligned knowledge information.

[0103] Optionally, the review construction module 304 may also be specifically used for:

[0104] Based on the preset historical knowledge graph library and the fused knowledge information, similarity calculation is performed to obtain the similarity of each entity sequence in the fused knowledge information;

[0105] Determine a target entity sequence by using a preset threshold and the similarity of each entity sequence, wherein the target entity sequence is an entity sequence whose similarity is less than or equal to the preset threshold;

[0106] Pre-audit the target entity sequence;

[0107] If the target entity sequences all pass the preliminary review, the target historical and cultural knowledge fusion model is constructed based on the fused knowledge information.

[0108] Optionally, the review construction module 304 may also be specifically used for:

[0109] If the target entity sequence includes an entity sequence that fails the pre-review, the entity sequence that fails the pre-review is trimmed to obtain a trimmed entity sequence;

[0110] According to the preset number of audits, the trimmed entity sequence is audited to obtain the target trimmed entity sequence;

[0111] Based on the target trimmed entity sequence, the entity sequence that has passed the pre-examination and the entity sequence whose similarity is greater than the preset threshold, a target historical and cultural knowledge fusion model is constructed.

[0112] The functional implementation of each module and each unit in the above-mentioned historical and cultural knowledge fusion model construction device corresponds to the various steps in the above-mentioned historical and cultural knowledge fusion model construction method embodiment, and their functions and implementation processes will not be repeated here one by one.

[0113] This embodiment also provides an electronic device, including a processor and a memory, wherein the memory stores machine executable instructions that can be executed by the processor, and the processor executes the machine executable instructions to implement the method for constructing the historical and cultural knowledge fusion model. The electronic device can be a server or a terminal device.

[0114] See also Figure 4 As shown, the electronic device includes a processor 400 and a memory 401, wherein the memory 401 stores machine executable instructions that can be executed by the processor 400, and the processor 400 executes the machine executable instructions to implement the above-mentioned method for constructing the historical and cultural knowledge fusion model.

[0115] Further, Figure 4 The electronic device shown further includes a bus 402 and a communication interface 403 , and the processor 400 , the communication interface 403 and the memory 401 are connected via the bus 402 .

[0116] The memory 401 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk storage. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 403 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used. The bus 402 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0117] The processor 400 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 400. The above processor 400 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The various methods, steps and logic block diagrams of the invention in the embodiment of the present invention can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method invented in conjunction with the embodiment of the present invention can be directly embodied as a hardware decoding processor for execution, or a combination of hardware and software modules in the decoding processor for execution. The software module may be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, or other mature storage media in the art.

[0118] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program runs on a computer, the computer executes the steps of the method for constructing a historical and cultural knowledge fusion model.

[0119] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0120] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several computer programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a historical and cultural knowledge fusion model, characterized in that: The method for constructing the historical and cultural knowledge fusion model includes: Acquire knowledge information of target historical and cultural field data after knowledge extraction, and determine the subject concept information of historical and cultural knowledge fusion, wherein the target historical and cultural field data is historical and cultural field data from multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes subject field and concept type; Perform entity alignment based on the subject concept information and the knowledge information to obtain aligned knowledge information; Performing multi-level fusion on the aligned knowledge information to obtain fused knowledge information; The integrated knowledge information is reviewed, and a target historical and cultural knowledge fusion model is constructed based on the reviewed integrated knowledge information.

2. The method for constructing a historical and cultural knowledge fusion model according to claim 1 is characterized in that: The aligned knowledge information is fused at multiple levels to obtain fused knowledge information, including: Performing attribute fusion on the aligned knowledge information to obtain initial fused knowledge information; Performing relational fusion on the initial fused knowledge information to obtain re-fused knowledge information; Based on preset historical and cultural field rules, the re-integrated knowledge information is integrated to obtain integrated knowledge information.

3. The method for constructing a historical and cultural knowledge fusion model according to claim 2 is characterized in that: The aligned knowledge information is subjected to attribute fusion to obtain initial fused knowledge information, including: Dividing the attributes in the aligned knowledge information according to the aligned entities to obtain the attributes of each similar entity; Classifying the attributes of the similar entities to obtain various attributes of the similar entities; Based on the preset historical and cultural field attribute strategy, various attributes of the similar entities are integrated to obtain initial fusion knowledge information.

4. The method for constructing a historical and cultural knowledge fusion model according to claim 2 is characterized in that: The performing relation fusion on the initial fused knowledge information to obtain the re-fused knowledge information includes: Classifying the entity relationships in the initial fused knowledge information to obtain classified entity relationships; Based on a preset historical and cultural field relationship template, the classified entity relationships are transformed to obtain integrated entity relationships; Based on the preset historical and cultural field relationship paths and the integrated entity relationships, the initial fusion knowledge information is relationally connected to obtain a historical and cultural relationship network; Based on a preset optimization algorithm, the historical and cultural relationship network is optimized to obtain the re-integrated knowledge information.

5. The method for constructing a historical and cultural knowledge fusion model according to claim 1 is characterized in that: The performing entity alignment based on the subject concept information and the knowledge information to obtain aligned knowledge information includes: Classifying the knowledge information based on the subject concept information to obtain multiple categories of information; Determine a target individual corresponding to each type of information in the multiple types of information based on a preset genetic algorithm, wherein the target individual indicates an optimal entity alignment solution; Based on the target individuals corresponding to each type of information, the corresponding entities are integrated to obtain aligned knowledge information.

6. The method for constructing a historical and cultural knowledge fusion model according to any one of claims 1 to 5, characterized in that: The step of reviewing the integrated knowledge information and constructing a target historical and cultural knowledge fusion model based on the reviewed integrated knowledge information includes: Perform similarity calculation based on a preset historical knowledge graph library and the fused knowledge information to obtain the similarity of each entity sequence in the fused knowledge information; Determining a target entity sequence by a preset threshold and the similarity of each entity sequence, wherein the target entity sequence is an entity sequence whose similarity is less than or equal to the preset threshold; Pre-auditing the target entity sequence; If the target entity sequences are all pre-reviewed and passed, a target historical and cultural knowledge fusion model is constructed based on the fused knowledge information.

7. The method for constructing a historical and cultural knowledge fusion model according to claim 6 is characterized in that: After the target entity sequence is pre-audited, the method further includes: If the target entity sequence includes an entity sequence that fails the pre-review, trimming the entity sequence that fails the pre-review to obtain a trimmed entity sequence; Reviewing the trimmed entity sequence according to a preset number of reviews to obtain a target trimmed entity sequence; Based on the target trimmed entity sequence, the entity sequence that has passed the pre-examination and the entity sequence whose similarity is greater than a preset threshold, a target historical and cultural knowledge fusion model is constructed.

8. A device for constructing a historical and cultural knowledge fusion model, characterized in that: The construction device of the historical and cultural knowledge fusion model includes: A determination module is used to obtain knowledge information of the target historical and cultural field data after knowledge extraction, and to determine the subject concept information of the historical and cultural knowledge fusion, wherein the target historical and cultural field data is the historical and cultural field data of multiple data sources that have been normalized by a preset multimodal model, and the subject concept information includes a subject field and a concept type; An alignment module, used for performing entity alignment based on the subject concept information and the knowledge information to obtain aligned knowledge information; A fusion module, used for performing multi-level fusion on the aligned knowledge information to obtain fused knowledge information; The review and construction module is used to review the integrated knowledge information and construct a target historical and cultural knowledge fusion model based on the integrated knowledge information that has passed the review.

9. An electronic device, characterized in that: The electronic device comprises: a memory and at least one processor, wherein the memory stores a computer program; The at least one processor calls the computer program in the memory so that the electronic device executes the method for constructing a historical and cultural knowledge fusion model as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the method for constructing a historical and cultural knowledge fusion model as described in any one of claims 1-7.

Images