A method and system for associative indexing based on graph database

Abstract

The present application provides an association indexing method and system based on a graph database, which relates to the field of knowledge retrieval technology. The method includes: traversing a knowledge file set to extract triples; building a scientific research knowledge graph based on an entity information set, a relationship information set, and an attribute information set; receiving user index request information; performing content association analysis to obtain a content association degree set; evaluating the value of files based on value attribute feature values ​​and the scientific research knowledge graph to obtain a file importance degree set; sorting the knowledge file set; and sending the index target file set to a user-side display interface for display. This application can solve the technical problem in the prior art that the search results have a low degree of matching with user needs due to insufficient analysis of multi-dimensional elements in semantic analysis, thereby achieving the technical goal of obtaining knowledge files that have a high degree of matching with user needs and achieving the technical effect of improving the efficiency of users in obtaining accurate search results.

Description

Technical Field

[0001] The present application relates to the field of knowledge retrieval technology, and in particular to an association indexing method and system based on a graph database. Background Art

[0002] In the scientific research process, knowledge retrieval is a very important part. Knowledge retrieval can help researchers quickly find relevant literature, data, experimental results, etc., thereby accelerating the scientific research process and improving research efficiency.

[0003] At present, most existing scientific research knowledge retrieval relies on simple machine learning for semantic recognition, which makes it difficult to consider multi-dimensional factors in the semantic analysis of the model. At the same time, the generalization ability is weak, resulting in the retrieval results being less applicable to user needs.

[0004] To sum up, there are technical problems in the existing technology that exist due to insufficient analysis of multi-dimensional factors in semantic analysis, resulting in a low degree of matching of retrieval results with user needs, reducing the efficiency of users in obtaining accurate retrieval results, leading to a decline in research quality, and further affecting the research process. Summary of the Invention

[0005] The purpose of this application is to provide an associative indexing method and system based on a graph database, so as to solve the technical problems in the prior art that, due to insufficient analysis of multi-dimensional elements in semantic analysis, the retrieval results have a low degree of matching with user needs, resulting in reduced efficiency for users to obtain accurate retrieval results, resulting in a decline in research quality, and further affecting the research process.

[0006] In view of the above problems, the present application provides an association indexing method and system based on a graph database.

[0007] In the first aspect, the present application provides an association indexing method based on a graph database, which is implemented by an association indexing system based on a graph database, wherein the method includes: traversing the knowledge file set to extract triples to obtain an entity information set, a relationship information set and an attribute information set; building a scientific research knowledge graph based on the entity information set, the relationship information set and the attribute information set; receiving user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values; performing content association analysis based on the scientific research knowledge graph based on the content attribute characteristic values ​​to obtain a content association set; performing file value evaluation based on the scientific research knowledge graph based on the value attribute characteristic values ​​to obtain a file importance set; sorting the knowledge file set based on the content association set and the file importance set to obtain an index target file set; and sending the index target file set to the user-end display interface for display.

[0008] In the second aspect, the present application also provides an association indexing system based on a graph database, which is used to execute an association indexing method based on a graph database as described in the first aspect, wherein the system includes: a triple acquisition module, the triple acquisition module is used to traverse the knowledge file set to extract triples, and obtain an entity information set, a relationship information set and an attribute information set; a scientific research knowledge graph building module, the scientific research knowledge graph building module is used to build a scientific research knowledge graph according to the entity information set, the relationship information set and the attribute information set; a user index request information receiving module, the user index request information receiving module is used to receive user index request information, wherein the user index request information includes content attribute feature values ​​and value attribute feature values; a content relevance set An acquisition module, the content relevance set acquisition module is used to perform content relevance analysis based on the scientific research knowledge graph according to the content attribute characteristic values, and obtain a content relevance set; a file importance set acquisition module, the file importance set acquisition module is used to perform file value evaluation based on the scientific research knowledge graph according to the value attribute characteristic values, and obtain a file importance set; an index target file set acquisition module, the index target file set acquisition module is used to sort the knowledge file set according to the content relevance set and the file importance set, and obtain an index target file set; an index target file set acquisition module, the index target file set acquisition module is used to send the index target file set to a user-end display interface for display.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] 1. Extract triples by traversing the knowledge file set to obtain an entity information set, a relationship information set and an attribute information set; build a scientific research knowledge graph based on the entity information set, the relationship information set and the attribute information set; receive user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values; perform content association analysis based on the scientific research knowledge graph based on the content attribute characteristic values ​​to obtain a content association set; perform file value evaluation based on the scientific research knowledge graph based on the value attribute characteristic values ​​to obtain a file importance set; sort the knowledge file set based on the content association set and the file importance set to obtain an index target file set; send the index target file set to the user-end display interface for display, that is, through the knowledge graph, perform dual association analysis of content and value on knowledge, and ultimately achieve the technical goal of obtaining knowledge files that highly match user needs, and achieve the technical effect of improving the efficiency of users in obtaining accurate retrieval results.

[0011] 2. This application is applicable to other knowledge retrieval scenarios and has the advantages of high generalization and strong scenario migration capabilities.

[0012] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically listed below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in this application or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and a person of ordinary skill in the art can obtain other drawings based on the provided drawings without creative work.

[0014] Figure 1 This is a flowchart of an association indexing method based on a graph database in this application;

[0015] Figure 2 This is a structural diagram of an association index system based on a graph database in this application.

[0016] Description of reference numerals:

[0017] Triple acquisition module 11, scientific research knowledge graph construction module 12, user index request information receiving module 13, content relevance set acquisition module 14, file importance set acquisition module 15, index target file set acquisition module 16, index target file set acquisition module 17. DETAILED DESCRIPTION

[0018] This application provides a graph database-based associative indexing method and system to address the existing technical issues of insufficient semantic analysis of multi-dimensional elements, which results in poor matching of search results with user needs. This reduces the efficiency of users in obtaining accurate search results, reduces research quality, and further affects the research process. This achieves the technical goal of obtaining knowledge documents that are highly compatible with user needs, achieving the technical effect of improving the efficiency of users in obtaining accurate search results.

[0019] Below, the technical solutions in this application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this application, rather than all the embodiments of this application. It should be understood that this application is not limited to the example embodiments described herein. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should also be noted that, for the convenience of description, only the parts related to this application, rather than all of them, are shown in the accompanying drawings.

[0020] For example 1, please refer to the attached Figure 1 The present application provides a graph database-based associative indexing method, wherein the method is applied to a graph database-based associative indexing system, and the method specifically comprises the following steps:

[0021] Step 1: Traverse the knowledge document set to extract triples and obtain entity information set, relationship information set and attribute information set;

[0022] Specifically, a knowledge document set is a collection of knowledge documents to be indexed, i.e., a knowledge document database. The knowledge document set is accessed sequentially, and triples are extracted from the knowledge document set for data indexing. Triples comprise an entity information set, a relationship information set, and an attribute information set. The entity information set is the content information set of the knowledge document. The relationship information set is the value type information set of the knowledge document. The attribute information set is the value evaluation attribute information set of the knowledge document.

[0023] Step 2: Building a scientific research knowledge graph based on the entity information set, the relationship information set, and the attribute information set;

[0024] Specifically, the entity information set, relationship information set and attribute information set in the triple are converted into nodes in the graph database, and then a scientific research knowledge graph is obtained based on multiple nodes.

[0025] Step 3: receiving user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values;

[0026] Specifically, the user index request information includes content attribute feature values ​​and value attribute feature values. Content attribute feature values ​​are digital features of the content being indexed. Value attribute feature values ​​are digital features of the value being indexed. Furthermore, the user index request information is received for subsequent associated indexing.

[0027] Step 4: performing content association analysis based on the scientific research knowledge graph according to the content attribute feature values ​​to obtain a content association degree set;

[0028] Specifically, the similarity between the attributes to be retrieved and the feature values ​​of the content attributes is analyzed for clustering to obtain a content relevance set. Then, when semantically associating with the feature values ​​of the content attributes in the subsequent process, only one of each category needs to be matched for representative calculation, thereby improving computational efficiency.

[0029] Step 5: Based on the value attribute characteristic value and the scientific research knowledge graph, the document value is evaluated to obtain a document importance set;

[0030] Specifically, we obtain the value attribute feature values ​​of the nodes in the scientific research knowledge graph, evaluate the value of the files based on the value attribute feature values, and obtain a set of file importance. The better the value attribute feature values, the higher the file importance, and vice versa.

[0031] Step 6: sorting the knowledge document set according to the content relevance set and the document importance set to obtain an index target document set;

[0032] Specifically, the index target files corresponding to the content relevance set are obtained, the files with low content relevance are deleted from the obtained index target files, the file importance is extracted from the deleted index target files, the files with low file importance are deleted from the obtained index target files, and the index target file set is obtained.

[0033] Step 7: Send the index target file set to the user terminal display interface for display.

[0034] Specifically, the multiple clusters of index target files are sorted according to their corresponding content relevance, and then sorted according to their corresponding file importance to obtain an index target file set, which is sent to a user terminal display interface for display.

[0035] The described associative indexing method based on a graph database is applied to an associative indexing system based on a graph database, which can extract triples by traversing a knowledge file set to obtain an entity information set, a relationship information set and an attribute information set; build a scientific research knowledge graph based on the entity information set, the relationship information set and the attribute information set; receive user index request information, wherein the user index request information includes a content attribute characteristic value and a value attribute characteristic value; perform content association analysis based on the scientific research knowledge graph based on the content attribute characteristic value to obtain a content association set; perform file value evaluation based on the scientific research knowledge graph based on the value attribute characteristic value to obtain a file importance set; sort the knowledge file set based on the content association set and the file importance set to obtain an index target file set; send the index target file set to a user-end display interface for display, that is, through the knowledge graph, perform a dual association analysis of content and value on knowledge, and ultimately achieve the technical goal of obtaining knowledge files that are highly matched with user needs, and achieve the technical effect of improving the efficiency of users in obtaining accurate retrieval results.

[0036] Furthermore, this application also includes:

[0037] Configuring an entity type set, wherein the entity type set at least includes scientific researchers, scientific research institutions, knowledge topics, and knowledge documents;

[0038] Configuring a relationship type set, wherein the relationship type set includes at least keywords, research methods, research data, citation times, journal impact factor, and author scientific value;

[0039] Configuring an attribute type set, wherein the attribute type set includes at least a keyword characteristic value, a research method characteristic value, a research data characteristic value, a citation count characteristic value, a journal impact factor characteristic value, and an author's scientific value characteristic value;

[0040] According to the entity type set, the relationship type set, and the attribute type set, the knowledge document set is traversed to perform triple extraction to obtain the entity information set, the relationship information set, and the attribute information set.

[0041] Specifically, the entity type set in the knowledge file set is configured, wherein the entity type set obtained by the configuration includes at least scientific researchers, scientific research institutions, knowledge topics and knowledge files. The entity type set may also include other scientific research knowledge types and contents. Furthermore, the entity type set is the scientific research knowledge type to be associated and indexed. Among them, scientific researchers are people who have scientific theoretical knowledge and are engaged in scientific research. Scientific research institutions are institutions with clear research directions and tasks, academic leaders and research teams of a certain level, and the basic conditions for conducting research. Knowledge topics are the core concepts, theories and knowledge systems of specific fields or topics. Knowledge files are files or documents that record, organize and store knowledge.

[0042] Then, the relationship type set in the knowledge document set is configured, wherein the relationship type set obtained by the configuration includes at least keywords, research methods, research data, citation times, journal impact factors, and author scientific value. The relationship type set can also include other scientific research knowledge value types. Furthermore, the relationship type set is the value of scientific research knowledge to be associated indexed. Among them, keywords are words or phrases used in the text to express themes, concepts or key points. Research methods are the methods and tools used in the research process. Research data are various data collected, generated and used in the scientific research process. Citation times are the number of times the knowledge content is cited in the references of other academic works. The journal impact factor is an indicator of the usefulness and visibility of the journal, and is also an important indicator of the academic level of the journal and even the quality of the paper. The author's scientific value is the contribution and influence made by the author in the field of scientific research.

[0043] Secondly, the attribute type set in the knowledge document set is configured, wherein the configured attribute type set includes at least keyword characteristic values, research method characteristic values, research data characteristic values, citation count characteristic values, journal impact factor characteristic values, and author scientific value characteristic values. The attribute type set may also include other scientific research knowledge attribute types. Furthermore, the attribute type set is the attribute of the scientific research knowledge to be associated and indexed. Among them, the keyword characteristic value is a numerical indicator used to describe and distinguish keywords. The research method characteristic value is a numerical indicator of the specific steps of the research. The research data characteristic value is a numerical indicator used to describe and study the characteristics of the data set. The citation count characteristic value is a numerical indicator used to describe and evaluate the characteristics related to the number of citations of academic works. The journal impact factor characteristic value is a numerical indicator for evaluating the academic level, reputation, influence, and paper quality of the journal. The author scientific value characteristic value is a numerical indicator used to describe and evaluate the value and influence demonstrated by the author in the field of scientific research.

[0044] In addition, the data included in the entity type set is sequentially accessed and combined to obtain the entity type set, the data included in the relationship type set is sequentially accessed and combined to obtain the relationship type set, and the data included in the attribute type set is sequentially accessed and combined to obtain the attribute type set.

[0045] By configuring the triples of the knowledge document set, the accuracy of the association indexing through the triples is improved.

[0046] Furthermore, this application also includes:

[0047] Retrieve knowledge document publication records based on the knowledge document author and retrieval time zone;

[0048] Traversing the publication records of the knowledge document, extracting a set of citation times and a set of journal impact factors;

[0049] Deleting outliers from the citation count set to obtain a citation count concentration value, calculating a mean of the citation count concentration value and performing normalization processing to set the mean as a citation count concentration evaluation value, wherein the citation count concentration evaluation value has a first preset weight;

[0050] Deleting outlier data from the journal impact factor set to obtain a concentrated impact factor value, calculating a mean of the concentrated impact factor value and performing normalization processing to set the mean as an impact factor concentrated evaluation value, wherein the concentrated impact factor evaluation value has a second preset weight, the second preset weight is greater than the first preset weight, and the sum of the second preset weight and the first preset weight is equal to one;

[0051] According to the second preset weight and the first preset weight, a weighted average evaluation is performed on the centralized evaluation value of the number of citations and the centralized evaluation value of the impact factor to generate the scientific value of the author.

[0052] Specifically, a knowledge document includes its author, who is the person who created or summarized the knowledge document. The search time zone is the time period for the search, for example, January 2024 to February 2024. Within the search time zone, searches are performed on the knowledge document author to obtain the knowledge document and its corresponding publication record. The knowledge document publication record is a record of the knowledge document's publication, including, for example, the publication date.

[0053] Then, each knowledge document publication record in the knowledge document publication record is accessed in turn, each citation count is extracted from the relationship type set corresponding to each knowledge document publication record and combined to obtain a citation count set, and each journal impact factor is extracted and combined to obtain a journal impact factor set.

[0054] Secondly, by drawing a histogram of the data, outlier data is observed in the citation count set to obtain outliers in the citation count set. The outliers are then removed from the citation count set using data interpolation to obtain the citation count concentration value. The mean of the citation count concentration value is calculated and normalized to obtain the data integrity of the citation count set. The normalized result is then set as the citation count concentration evaluation value, which is used to conduct subsequent scientific value assessments of the authors. The citation count concentration evaluation value has a first preset weight. The more citations, the higher the citation count concentration evaluation value, and the higher the first preset weight, and vice versa.

[0055] In addition, by plotting the data histogram, we observed outliers in the journal impact factor set and obtained the outliers in the journal impact factor set. We then used data interpolation to remove the outliers from the journal impact factor set and obtain the impact factor concentration value. By calculating the mean of the impact factor concentration value and then normalizing it, we can ensure the data integrity of the journal impact factor set. The normalized result is then set as the impact factor concentration evaluation value, which is used to conduct subsequent scientific value assessments of authors. The impact factor concentration evaluation value has a second preset weight. When the reputation and influence of the journal are higher, the impact factor concentration evaluation value is higher, and the second preset weight is higher. Conversely, the second preset weight is lower.

[0056] Furthermore, because the importance of the journal impact factor set is greater than the citation count set, the second preset weight is greater than the first preset weight, and the sum of the second preset weight and the first preset weight is equal to one. For example, the ratio of the second preset weight to the first preset weight is 0.6:0.4, and the sum of the second preset weight and the first preset weight is one.

[0057] In addition, the first preset weight is multiplied by the centralized evaluation value of the number of citations, and the second preset weight is multiplied by the centralized evaluation value of the impact factor. The two product results are added together to obtain the author's scientific value.

[0058] The author's scientific value is obtained by calculating the centralized evaluation value of citation times and the centralized evaluation value of impact factor, thereby improving the accuracy of obtaining the author's scientific value.

[0059] Furthermore, the present application further comprises the following steps:

[0060] The content attribute feature value includes at least the keyword feature value, the research method feature value, and the research data feature value;

[0061] The value attribute characteristic values ​​include at least the citation times characteristic value, the journal impact factor characteristic value, and the author scientific value characteristic value.

[0062] Specifically, the content attribute feature value includes at least the keyword feature value, research method feature value and research data feature value of the attribute type set. Furthermore, the content of the knowledge document is evaluated based on the keyword feature value, research method feature value and research data feature value.

[0063] Then, the value attribute characteristic value includes at least the attribute type set of the number of citations characteristic value, the journal impact factor characteristic value and the author's scientific value characteristic value. Further, the value of the knowledge document is evaluated by the number of citations characteristic value, the journal impact factor characteristic value and the author's scientific value characteristic value.

[0064] By obtaining the specific characteristic values ​​of the content attribute characteristic values ​​and the value attribute characteristic values, the accuracy of subsequent analysis of user index request information can be improved.

[0065] Furthermore, the present application further comprises the following steps:

[0066] According to the scientific research knowledge graph, a first attribute information set, a second attribute information set, and so on to an Nth attribute information set are obtained, where N is an integer and N≥1;

[0067] Performing semantic clustering analysis on the first attribute information set, the second attribute information set, up to the Nth attribute information set, to obtain first cluster attribute information, second cluster attribute information, up to the Mth cluster attribute information, where M is an integer and M≥1;

[0068] The first cluster attribute information, the second cluster attribute information, and the Mth cluster attribute information are traversed, and semantic similarity evaluation is performed with the content attribute feature value to obtain M semantic similarity coefficients, which are set as the content association degree set.

[0069] Specifically, attribute information sets are randomly extracted from the constructed scientific research knowledge graph to obtain the first attribute information set, the second attribute information set, and so on to the Nth attribute information set, and the scientific research knowledge graph is accessed in sequence to obtain all attribute information sets, wherein, since the scientific research knowledge graph includes at least one attribute information set, N is an integer, and N≥1.

[0070] Then, semantic clustering analysis is performed on the first attribute information set, the second attribute information set, and so on, up to the Nth attribute information set. That is, the similarities between the attributes to be retrieved are analyzed and clustered. Then, when semantically associating with the content attribute feature values, only one attribute information set of each category needs to be matched for representative calculation, thereby improving computational efficiency. Semantic clustering analysis is performed to obtain the first cluster attribute information, the second cluster attribute information, and so on, up to the Mth cluster attribute information, where M is an integer, M ≥ 1. The values ​​of N and M are not equal. For example, the number of clusters of attribute information obtained after clustering may be greater than, equal to, or less than the number of attribute information sets.

[0071] Next, the first cluster of attribute information, then the second cluster, and finally the Mth cluster are accessed sequentially. Semantic similarity is assessed with the content attribute feature values ​​for semantic association. This allows for only one attribute information set per category to be matched for representative computation, improving computational efficiency. M semantic similarity coefficients are obtained through the semantic similarity assessment and are defined as the content association set. A higher semantic similarity coefficient indicates a higher content association, and vice versa.

[0072] A content relevance set is obtained through content relevance analysis, thereby improving the accuracy of the obtained content relevance set.

[0073] Furthermore, the present application further comprises the following steps:

[0074] The value attribute characteristic value includes several value attributes;

[0075] Traversing the plurality of value attributes to perform weight distribution and obtain a value attribute weight distribution result;

[0076] According to the value attribute weight distribution results, a file importance evaluation function is constructed:

[0077]

[0078] Among them, z k Characterize the document importance of the k-th value attribute information set of the scientific research knowledge graph, x ki Represents the i-th value attribute record value of the k-th value attribute information set, Q represents the total number of value attributes, x 0i The value corresponding to the i-th value attribute characteristic value, w i The weight distribution result representing the i-th value attribute;

[0079] According to the file importance evaluation function, the file value evaluation is performed based on the scientific research knowledge graph to obtain the file importance set.

[0080] Specifically, since the knowledge document set has multiple value attributes, the value attribute characteristic values ​​are obtained, including several value attributes. For example, the value attributes include the characteristic value of the number of citations, the characteristic value of the journal impact factor, and the characteristic value of the author's scientific value.

[0081] Then, we sequentially access several value attributes and perform weight distribution. The better the value attribute, the greater the weight of the obtained value attribute, and vice versa. For example, when the number of citations of the knowledge document is higher, the journal impact factor is higher, and the author's scientific value is higher, the value attribute is better, and the weight of the obtained value attribute is set to be larger.

[0082] Secondly, based on the value attribute weight distribution results, a file importance evaluation function is constructed:

[0083]

[0084] Among them, x ki Characterize the i-th value attribute record value of the k-th value attribute information set. For example, when the k-th value attribute information set is the characteristic value of the number of citations, the i-th value attribute record value can be twice, so x is obtained. ki The value of Q is 2. Q represents the total number of value attributes. For example, if the value attributes include the characteristic value of the number of citations, the characteristic value of the journal impact factor, and the characteristic value of the author's scientific value, the total number of value attributes is 3. 0i The value corresponding to the i-th value attribute characteristic value representing the value attribute characteristic value. For example, the value attribute characteristic value of the number of citations, the journal impact factor characteristic value, and the author's scientific value characteristic value are obtained. The i-th value attribute characteristic value can be 2 for the number of citations, 3 for the journal impact factor, 4 for the author's scientific value, etc. i The weight distribution result representing the i-th value attribute. The higher the i-th value attribute, the higher the weight distribution result w i The larger the value, the lower the value. For example, the higher the eigenvalue of the number of citations, the weight distribution result w i The larger the value, the larger the value. k Characterizes the file importance of the k-th value attribute information set of the scientific research knowledge graph, where the file importance of the k-th value attribute information set is obtained by the weight ratio of the k-th value attribute characteristic value to the value attribute characteristic value. When the weight ratio of the k-th value attribute characteristic value to the value attribute characteristic value is higher, the file importance of the k-th value attribute information set is higher, and vice versa.

[0085] In addition, the file importance evaluation function is input based on the corresponding value attribute characteristic values ​​in the scientific research knowledge graph to evaluate the file value, and then multiple file importances are obtained, and the file importance set is obtained by combination.

[0086] The file importance is obtained by evaluating the value attribute characteristic value through the file importance evaluation function, thereby improving the accuracy of the obtained file importance.

[0087] Furthermore, the present application further comprises the following steps:

[0088] The content relevance set has a one-to-one correspondence of multiple cluster index target files;

[0089] Deleting clusters whose content relevance is less than or equal to a content relevance threshold from the multi-cluster index target file to obtain a first-level sorting result of the index target file;

[0090] According to the file importance set, files that are less than or equal to a file importance threshold are deleted from the first-level sorting result of the index target files to obtain the index target file set.

[0091] Specifically, the content relevance set has a one-to-one correspondence with a multi-cluster index target file, wherein attribute information sets are randomly extracted through the scientific research knowledge graph to obtain multiple attribute information sets. Similarity clustering is performed on the attribute information sets to obtain multi-cluster attribute information, and the corresponding multi-cluster index target file is obtained based on the multi-cluster attribute information.

[0092] Next, a content relevance threshold is obtained. This threshold can be customized by those skilled in the art based on practical circumstances, for example, 80%. If the content relevance in the obtained content relevance set is less than or equal to the content relevance threshold, indicating that the corresponding index target file has a low content relevance, the corresponding cluster of the index target file is deleted from the multi-cluster index target file, obtaining a first-level sorting result for the index target file. For example, the remaining multi-cluster index target files after deletion are subjected to mean interpolation to obtain a first-level sorting result for the index target file, thereby preserving data integrity.

[0093] Next, a file importance threshold is obtained. This threshold can be customized by those skilled in the art based on actual circumstances, for example, 80%. Files whose file importance is less than or equal to the threshold are removed from the primary sorting results of the index target files, thereby obtaining an index target file set. For example, the remaining index target files after deletion are subjected to mean interpolation to obtain an index target file set, thereby preserving data integrity.

[0094] The index target file set is obtained by performing multi-level deletion processing on the multi-cluster index target files, thereby improving the simplicity of the obtained index target file set.

[0095] Furthermore, the present application further comprises the following steps:

[0096] The content relevance set has a one-to-one correspondence of multiple cluster index target files;

[0097] Sorting the multi-cluster index target files according to the content relevance set from largest to smallest to obtain a multi-cluster index target file sorting result;

[0098] According to the file importance set, the multi-cluster index target file sorting results are traversed and sorted from large to small to obtain the index target file sorting results;

[0099] The index target file set is sent to a user terminal display interface for display according to the index target file sorting result.

[0100] Specifically, the content relevance set has a one-to-one correspondence with a multi-cluster index target file. Attribute information sets are randomly extracted from the scientific research knowledge graph to obtain multiple attribute information sets. Similarity clustering is performed on the attribute information sets to obtain multi-cluster attribute information, and corresponding multi-cluster index target files are obtained based on the multi-cluster attribute information.

[0101] Then, the content relevance sets corresponding to the multi-cluster index target files are sorted from large to small according to the content relevance to obtain the content relevance sorting result, and then the corresponding multi-cluster index target file sorting result is obtained as the multi-cluster index target file sorting result.

[0102] Secondly, extract the file importance set corresponding to the multi-cluster index target files in the multi-cluster index target file sorting result, sort the file importance in the file importance set from large to small, obtain the file importance sorting result, and then obtain the corresponding multi-cluster index target file sorting result as the index target file sorting result.

[0103] In addition, the index target files in the index target file sorting result are combined according to the index target file sorting result to obtain an index target file set which is sent to the user terminal display interface for display.

[0104] By sorting the multi-cluster index target files according to the corresponding content relevance and then sorting them according to the corresponding file importance, an index target file set is obtained, thereby improving the neatness of the obtained index target file set.

[0105] In summary, the graph database-based associative indexing method provided by this application has the following technical effects:

[0106] 1. Extract triples by traversing the knowledge file set to obtain an entity information set, a relationship information set and an attribute information set; build a scientific research knowledge graph based on the entity information set, the relationship information set and the attribute information set; receive user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values; perform content association analysis based on the scientific research knowledge graph based on the content attribute characteristic values ​​to obtain a content association set; perform file value evaluation based on the scientific research knowledge graph based on the value attribute characteristic values ​​to obtain a file importance set; sort the knowledge file set based on the content association set and the file importance set to obtain an index target file set; send the index target file set to the user-end display interface for display, that is, through the knowledge graph, perform dual association analysis of content and value on knowledge, and ultimately achieve the technical goal of obtaining knowledge files that highly match user needs, and achieve the technical effect of improving the efficiency of users in obtaining accurate retrieval results.

[0107] 2. This application is applicable to other knowledge retrieval scenarios and has the advantages of high generalization and strong scenario migration capabilities.

[0108] In the second embodiment, based on the association indexing method based on the graph database in the above embodiment, the same inventive concept is also provided in this application. Figure 2 , the system comprising:

[0109] A triple acquisition module 11, which is used to traverse the knowledge document set to extract triples and obtain an entity information set, a relationship information set, and an attribute information set;

[0110] A scientific research knowledge graph building module 12, wherein the scientific research knowledge graph building module 12 is used to build a scientific research knowledge graph based on the entity information set, the relationship information set, and the attribute information set;

[0111] A user index request information receiving module 13, wherein the user index request information receiving module 13 is used to receive user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values;

[0112] A content relevance set acquisition module 14 is configured to perform content relevance analysis based on the scientific research knowledge graph according to the content attribute feature values ​​to obtain a content relevance set;

[0113] A file importance set obtaining module 15, which is used to evaluate the value of files based on the scientific research knowledge graph according to the value attribute characteristic value, and obtain a file importance set;

[0114] An index target file set obtaining module 16 is configured to sort the knowledge file set according to the content relevance set and the file importance set to obtain an index target file set;

[0115] The index target file set obtaining module 17 is used to send the index target file set to the user terminal display interface for display.

[0116] Furthermore, the triplet obtaining module 11 in the system is further configured to:

[0117] Configuring an entity type set, wherein the entity type set at least includes scientific researchers, scientific research institutions, knowledge topics, and knowledge documents;

[0118] Configuring a relationship type set, wherein the relationship type set includes at least keywords, research methods, research data, citation times, journal impact factor, and author scientific value;

[0119] Configuring an attribute type set, wherein the attribute type set includes at least a keyword characteristic value, a research method characteristic value, a research data characteristic value, a citation count characteristic value, a journal impact factor characteristic value, and an author's scientific value characteristic value;

[0120] According to the entity type set, the relationship type set, and the attribute type set, the knowledge document set is traversed to perform triple extraction to obtain the entity information set, the relationship information set, and the attribute information set.

[0121] Furthermore, the triplet obtaining module 11 in the system is further configured to:

[0122] Retrieve knowledge document publication records based on the knowledge document author and retrieval time zone;

[0123] Traversing the publication records of the knowledge document, extracting a set of citation times and a set of journal impact factors;

[0124] Deleting outliers from the citation count set to obtain a citation count concentration value, calculating a mean of the citation count concentration value and performing normalization processing to set the mean as a citation count concentration evaluation value, wherein the citation count concentration evaluation value has a first preset weight;

[0125] Deleting outlier data from the journal impact factor set to obtain a concentrated impact factor value, calculating a mean of the concentrated impact factor value and performing normalization processing to set the mean as an impact factor concentrated evaluation value, wherein the concentrated impact factor evaluation value has a second preset weight, the second preset weight is greater than the first preset weight, and the sum of the second preset weight and the first preset weight is equal to one;

[0126] According to the second preset weight and the first preset weight, a weighted average evaluation is performed on the centralized evaluation value of the number of citations and the centralized evaluation value of the impact factor to generate the scientific value of the author.

[0127] Furthermore, the user index request information receiving module 13 in the system is further configured to:

[0128] The content attribute feature value includes at least the keyword feature value, the research method feature value, and the research data feature value;

[0129] The value attribute characteristic values ​​include at least the citation times characteristic value, the journal impact factor characteristic value, and the author scientific value characteristic value.

[0130] Furthermore, the content relevance set obtaining module 14 in the system is further configured to:

[0131] According to the scientific research knowledge graph, a first attribute information set, a second attribute information set, and so on to an Nth attribute information set are obtained, where N is an integer and N≥1;

[0132] Performing semantic clustering analysis on the first attribute information set, the second attribute information set, up to the Nth attribute information set, to obtain first cluster attribute information, second cluster attribute information, up to the Mth cluster attribute information, where M is an integer and M≥1;

[0133] The first cluster attribute information, the second cluster attribute information, and the Mth cluster attribute information are traversed, and semantic similarity evaluation is performed with the content attribute feature value to obtain M semantic similarity coefficients, which are set as the content association degree set.

[0134] Furthermore, the file importance set obtaining module 15 in the system is further configured to:

[0135] The value attribute characteristic value includes several value attributes;

[0136] Traversing the plurality of value attributes to perform weight distribution and obtain a value attribute weight distribution result;

[0137] According to the value attribute weight distribution results, a file importance evaluation function is constructed:

[0138]

[0139] Among them, z k Characterize the document importance of the k-th value attribute information set of the scientific research knowledge graph, x ki represents the i-th value attribute record value of the k-th value attribute information set, Q represents the total number of value attributes, x 0i The value corresponding to the i-th value attribute characteristic value, w i The weight distribution result representing the i-th value attribute;

[0140] According to the file importance evaluation function, the file value evaluation is performed based on the scientific research knowledge graph to obtain the file importance set.

[0141] Furthermore, the index target file set obtaining module 16 in the system is further configured to:

[0142] The content relevance set has a one-to-one correspondence of multiple cluster index target files;

[0143] Deleting clusters whose content relevance is less than or equal to a content relevance threshold from the multi-cluster index target file to obtain a first-level sorting result of the index target file;

[0144] According to the file importance set, files that are less than or equal to a file importance threshold are deleted from the first-level sorting result of the index target files to obtain the index target file set.

[0145] Furthermore, the index target file set obtaining module 17 in the system is further configured to:

[0146] The content relevance set has a one-to-one correspondence of multiple cluster index target files;

[0147] Sorting the multi-cluster index target files according to the content relevance set from largest to smallest to obtain a multi-cluster index target file sorting result;

[0148] According to the file importance set, the multi-cluster index target file sorting results are traversed and sorted from large to small to obtain the index target file sorting results;

[0149] The index target file set is sent to a user terminal display interface for display according to the index target file sorting result.

[0150] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The association indexing method based on a graph database and the specific examples in the aforementioned embodiment 1 are also applicable to the association indexing system based on a graph database in this embodiment. Through the aforementioned detailed description of the association indexing method based on a graph database, those skilled in the art can clearly understand the association indexing system based on a graph database in this embodiment. Therefore, for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple. For relevant details, please refer to the method description.

[0151] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0152] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application is intended to include these modifications and variations.

Claims

1. A graph database-based associative indexing method, characterized in that: include: Traverse the knowledge document set to extract triples and obtain entity information set, relationship information set and attribute information set; Building a scientific research knowledge graph based on the entity information set, the relationship information set, and the attribute information set; Receiving user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values; According to the content attribute feature values, content association analysis is performed based on the scientific research knowledge graph to obtain a content association degree set; According to the value attribute characteristic value, the document value is evaluated based on the scientific research knowledge graph to obtain a document importance set; Sorting the knowledge document set according to the content relevance set and the document importance set to obtain an index target document set includes: The value attribute characteristic value includes several value attributes; Traversing the plurality of value attributes to perform weight distribution and obtain a value attribute weight distribution result; According to the value attribute weight distribution results, a file importance evaluation function is constructed: Among them, z k Characterize the document importance of the k-th value attribute information set of the scientific research knowledge graph, x ki represents the i-th value attribute record value of the k-th value attribute information set, Q represents the total number of value attributes, x 0i The value corresponding to the i-th value attribute characteristic value, w i The weight distribution result representing the i-th value attribute; According to the file importance evaluation function, the file value evaluation is performed based on the scientific research knowledge graph to obtain the file importance set; The index target file set is sent to the user terminal display interface for display.

2. The method according to claim 1, wherein Traverse the knowledge document set to extract triples and obtain entity information set, relationship information set and attribute information set, including: Configuring an entity type set, wherein the entity type set at least includes scientific researchers, scientific research institutions, knowledge topics, and knowledge documents; Configuring a relationship type set, wherein the relationship type set includes at least keywords, research methods, research data, citation times, journal impact factor, and author scientific value; Configuring an attribute type set, wherein the attribute type set includes at least a keyword characteristic value, a research method characteristic value, a research data characteristic value, a citation count characteristic value, a journal impact factor characteristic value, and an author's scientific value characteristic value; According to the entity type set, the relationship type set, and the attribute type set, the knowledge document set is traversed to perform triple extraction to obtain the entity information set, the relationship information set, and the attribute information set.

3. The method according to claim 2, wherein The steps for assessing the scientific value of the author include: Retrieve knowledge document publication records based on the knowledge document author and retrieval time zone; Traversing the publication records of the knowledge document, extracting a set of citation times and a set of journal impact factors; Deleting outliers from the citation count set to obtain a citation count concentration value, calculating a mean of the citation count concentration value and performing normalization processing to set the mean as a citation count concentration evaluation value, wherein the citation count concentration evaluation value has a first preset weight; Deleting outlier data from the journal impact factor set to obtain a concentrated impact factor value, calculating a mean of the concentrated impact factor value and performing normalization processing to set the mean as an impact factor concentrated evaluation value, wherein the concentrated impact factor evaluation value has a second preset weight, the second preset weight is greater than the first preset weight, and the sum of the second preset weight and the first preset weight is equal to one; According to the second preset weight and the first preset weight, a weighted average evaluation is performed on the centralized evaluation value of the number of citations and the centralized evaluation value of the impact factor to generate the scientific value of the author.

4. The method according to claim 2, wherein Receive user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values, including: The content attribute feature value includes at least the keyword feature value, the research method feature value, and the research data feature value; The value attribute characteristic values ​​include at least the citation times characteristic value, the journal impact factor characteristic value, and the author scientific value characteristic value.

5. The method according to claim 1, wherein According to the content attribute feature values, content association analysis is performed based on the scientific research knowledge graph to obtain a content association degree set, including: According to the scientific research knowledge graph, a first attribute information set, a second attribute information set, and so on to an Nth attribute information set are obtained, where N is an integer and N≥1; Performing semantic clustering analysis on the first attribute information set, the second attribute information set, up to the Nth attribute information set, to obtain first cluster attribute information, second cluster attribute information, up to the Mth cluster attribute information, where M is an integer and M≥1; The first cluster attribute information, the second cluster attribute information, and the Mth cluster attribute information are traversed, and semantic similarity evaluation is performed with the content attribute feature value to obtain M semantic similarity coefficients, which are set as the content association degree set.

6. The method according to claim 1, wherein Sorting the knowledge document set according to the content relevance set and the document importance set to obtain an index target document set includes: The content relevance set has a one-to-one correspondence of multiple cluster index target files; Deleting clusters whose content relevance is less than or equal to a content relevance threshold from the multi-cluster index target file to obtain a first-level sorting result of the index target file; According to the file importance set, files that are less than or equal to a file importance threshold are deleted from the first-level sorting result of the index target files to obtain the index target file set.

7. The method according to claim 1, wherein Sending the index target file set to a user terminal display interface for display includes: The content relevance set has a one-to-one correspondence of multiple cluster index target files; sorting the multi-cluster index target files according to the content relevance set from largest to smallest to obtain a multi-cluster index target file sorting result; According to the file importance set, the multi-cluster index target file sorting results are traversed and sorted from large to small to obtain the index target file sorting results; The index target file set is sent to a user terminal display interface for display according to the index target file sorting result.

8. An association index system based on a graph database, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: A triple acquisition module, which is used to traverse the knowledge document set to extract triples and obtain entity information sets, relationship information sets, and attribute information sets; A scientific research knowledge graph building module, which is used to build a scientific research knowledge graph based on the entity information set, the relationship information set, and the attribute information set; A user index request information receiving module, the user index request information receiving module is used to receive user index request information, wherein the user index request information includes content attribute characteristic values ​​and value attribute characteristic values; A content relevance set acquisition module, wherein the content relevance set acquisition module is used to perform content relevance analysis based on the scientific research knowledge graph according to the content attribute feature values ​​to obtain a content relevance set; A file importance set acquisition module, which is used to evaluate the value of files based on the scientific research knowledge graph according to the value attribute characteristic value to obtain a file importance set; An index target file set obtaining module, configured to sort the knowledge file set according to the content relevance set and the file importance set to obtain an index target file set; The index target file set obtaining module is used to send the index target file set to the user terminal display interface for display.

Images