An information retrieval method based on knowledge graph

Through the information retrieval method based on knowledge graph, the Ripple Net model is used to process decision makers' questions, generate user vectors and predict probability, which solves the problem that traditional methods are difficult to obtain user entity interaction information, improves information retrieval efficiency, and helps decision makers quickly obtain Key information.

CN114691883BActive Publication Date: 2024-11-05NAT UNIV OF DEFENSE TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210223535.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-07
Publication Date
2024-11-05
Estimated Expiration
2042-03-07

AI Technical Summary

Technical Problem

Traditional information retrieval methods are difficult to obtain interactive information between user entities, ignore dynamic changes between users and multiple elements, making it difficult to efficiently provide key information for decision makers in the face of massive information.

Method used

Information retrieval method based on knowledge graphs, by constructing a decision-maker question information vocabulary, segmenting the question into element names and values, generating user vectors and inputting the Ripple Net model for iteration, forming an overall user vector representation, combining the target thing to predict the probability, Generate user search recommendation vector.

Benefits of technology

It realizes the connection between multi-factor node information, builds a multi-factor iteration cycle for decision makers user Q&A applications, improves information retrieval efficiency, and helps decision makers quickly obtain key information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114691883B_ABST
    Figure CN114691883B_ABST
Patent Text Reader

Abstract

The present invention provides an information retrieval method based on knowledge graph, which constructs a decision maker question information vocabulary according to different element attributes according to the concepts, entities and relationships in the decision maker's question, and divides the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, thereby constructing a question element model. On the basis of parsing the decision maker user's question, the association between the decision problem element entity and the knowledge base based on data is constructed to establish an iteratively expandable dynamic entity knowledge graph; a retrieval pair is constructed with the decision maker user identifier, the retrieval element name, and the entity name of the retrieval result, and the iterative algorithm in the Ripple Net method is used to realize the probability prediction of the response information, thereby realizing information response recommendation, solving the current information overload problem caused by massive information, and can help decision makers quickly obtain key information, and effectively improve the efficiency of information analysis and screening.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an information retrieval method based on a knowledge graph. Background Art

[0002] With the development of information technology and the Internet, people are gradually moving from a lack of information to an era of information overload. In order to solve this problem, the recommendation system came into being. The task of the recommendation system is to connect users and information, help users find valuable information for themselves, and improve the efficiency of users in obtaining key information.

[0003] At present, traditional methods mainly use explicit / implicit feedback interaction data as input to extract high- and low-order interaction features between users and items for recommendation, and on this basis, introduce large-scale node information in the knowledge graph and the attribute information of the nodes themselves. However, this method is difficult to obtain interaction information between user entities, and it is easy to ignore the dynamic changes between users and multiple factors. When faced with complex project information, it is difficult to efficiently provide decision makers with key information for retrieval;

[0004] Therefore, a method is needed to improve the efficiency of information retrieval to facilitate decision makers to make choices. Summary of the invention

[0005] The present invention provides an information retrieval method based on knowledge graph to solve the defects in the above-mentioned prior art, realize the connection of multi-factor node information, and construct a multi-factor iterative loop of decision maker retrieval questions in decision maker user question and answer application recommendation, so as to efficiently provide key information for decision makers.

[0006] The present invention provides an information retrieval method based on a knowledge graph, comprising the following steps:

[0007] S1. Construct the decision maker's question information word library according to different element attributes based on the concepts, entities and relationships in the decision maker's question;

[0008] S2. Segment the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, and form element pairs with element names and corresponding element values;

[0009] S3, extracting element values ​​from the plurality of element pairs in sequence according to the sentence order of the input decision maker's question, and then constructing a question element model;

[0010] S4, obtaining a decision maker user ID, obtaining the name of each search element extracted after preprocessing the decision maker's question, obtaining the entity name of each search result, constructing a search pair with the decision maker user ID, each search element name, and the entity name of each search result; and storing the search pair in a search record library;

[0011] S5. Generate a decision maker user vector based on the knowledge graph; iterate the decision maker user vector and the search pair into the Ripple Net model until the decision maker user vector representation of each link in the Ripple Net model is added to form an overall decision maker user vector representation;

[0012] S6. Combining the overall decision maker user vector representation and the target thing, predict the probability of the decision maker retrieving any item, and generate a decision maker user retrieval recommendation vector.

[0013] It should be noted that the Ripple Net model is a ripple preference propagation model, which is often used to propagate preference information between nodes based on the statistical data of user preferences for a certain thing, with the help of semantic relationships between nodes in the knowledge graph, and recommend retrieval conclusion information to users through entity nodes and semantic relationship features;

[0014] Further, in step S1, different element attributes include entity E, relationship R, attribute value V, concept C, space A, and time T respectively.

[0015] Further, in step S4, it includes:

[0016] The decision maker user ID U is retrieved, and the name of each search factor extracted after the decision maker's question preprocessing is F i , get the entity name of each search result as E j ;

[0017] The search pair S is constructed as: S: <U,F 1 , F 2 , ... F i , E 1 , E 2 ..., E j >.

[0018] Specifically, in step S5, based on the knowledge graph, a decision maker user vector is generated, including:

[0019] Vectorize the entities and relationships between entities in the decision maker's question to form a graph triple (h, r, t) covering all entities;

[0020] Among them, h represents the head entity, r represents the relationship, and t represents the tail entity.

[0021] Specifically, in step S6, the decision maker user vector and the search pair are input into Ripple Net for iteration, including:

[0022] Taking each of the retrieval pairs S as a node of the Ripple Net model, based on the knowledge graph G and the decision-maker user's question, the entity set of the decision-maker user in the kth round of iteration is:

[0023]

[0024] The set of triples for the kth iteration is defined as:

[0025]

[0026] When the Ripple Net model is substituted for the first iteration, the triplet of the first ring Ripple Set is obtained as follows:

[0027]

[0028] Based on each triple in the first ring Ripple Set, the vector R of the relationship i , vector h of the head entity i Multiply it with the vector v of the target thing to obtain the correlation relationship between the target thing v and the head entity and relationship in each triplet of the first iteration, thereby obtaining the decision maker user vector representation of the first-ring Ripple Set.

[0029] Specifically, in step S6, the decision maker user vector and the search pair are input into Ripple Net for iteration, further comprising:

[0030] When the Ripple Net model is substituted into the Ripple Net model for iteration for the kth time, the tail entity of the Ripple set in the k-1th iteration is used as the head entity of the kth iteration, thereby generating the user representation of the Ripple set in the kth iteration.

[0031] Preferably, after step S6, the method further comprises:

[0032] The decision maker user retrieval recommendation vector is input into a preset knowledge graph retrieval engine, and the entity name corresponding to the decision maker user retrieval recommendation vector is output.

[0033] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-described information retrieval methods are implemented.

[0034] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above-mentioned information retrieval methods are implemented.

[0035] The present invention provides an information retrieval method based on knowledge graph, which constructs a decision maker question information vocabulary according to different element attributes by respectively dividing the concepts, entities and relationships in the decision maker's question, and divides the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, and then constructs a question element model, which is conducive to building a data-based association between decision problem element entities and knowledge bases on the basis of parsing the decision maker user's question to establish an iteratively expandable dynamic entity knowledge graph, and covers the description of time and space elements; then, a retrieval pair is constructed with the decision maker user identifier, each retrieval element name, and the entity name of each retrieval result, and the retrieval pair is stored in a retrieval record library, and the iterative algorithm in the Ripple Net method is used to realize the probability prediction of the response information, thereby realizing information response recommendation, solving the current information overload problem caused by massive information, and can help decision makers quickly obtain key information, and effectively improve the efficiency of information analysis and screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the present invention 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 some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is one of the flow charts of the information retrieval method based on knowledge graph provided by the present invention;

[0038] Figure 2 This is the second flow chart of the information retrieval method based on knowledge graph provided by the present invention;

[0039] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0041] In one embodiment, Figure 1 As shown, the present invention provides an information retrieval method based on knowledge graph, comprising the following steps:

[0042] S1. Construct the decision maker's question information word library according to different element attributes based on the concepts, entities and relationships in the decision maker's question;

[0043] S2. Segment the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, and form element pairs with element names and corresponding element values;

[0044] S3, extracting element values ​​from the plurality of element pairs in sequence according to the sentence order of the input decision maker's question, and then constructing a question element model;

[0045] S4, obtaining a decision maker user ID, obtaining the name of each search element extracted after preprocessing the decision maker's question, obtaining the entity name of each search result, constructing a search pair with the decision maker user ID, each search element name, and the entity name of each search result; and storing the search pair in a search record library;

[0046] S5. Generate a decision maker user vector based on the knowledge graph; iterate the decision maker user vector and the search pair into the Ripple Net model until the decision maker user vector representation of each link in the Ripple Net model is added to form an overall decision maker user vector representation;

[0047] S6. Combining the overall decision maker user vector representation and the target thing, predict the probability of the decision maker retrieving any item, and generate a decision maker user retrieval recommendation vector.

[0048] It should be noted that the Ripple Net model is a ripple preference propagation model, which is often used to propagate preference information between nodes based on the statistical data of user preferences for a certain thing, with the help of semantic relationships between nodes in the knowledge graph, and recommend retrieval conclusion information to users through entity nodes and semantic relationship features;

[0049] Further, in step S1, different element attributes include six elements, namely entity E, relationship R, attribute value V, concept C, space A, and time T; wherein the decision maker question information word library is one of the decision maker question information association databases;

[0050] Specifically, an example of the element pair formed in step S2 is: <date, T>, <region, A>, <entity, E>, <behavior, R>, <question category, V>, <concept, C>;

[0051] Specifically, in step S3, the question element pattern constructed is: <t>| | <e> | <r> | <v> | <c>;

[0052] Further, in step S4, it includes:

[0053] The decision maker user ID U is retrieved, and the name of each search factor extracted after the decision maker's question preprocessing is F i , get the entity name of each search result as E j ;

[0054] The search pair S is constructed as: S: <U,F 1 , F 2 , ... F i , E 1 , E 2 ..., E j >.

[0055] Specifically, in step S5, based on the knowledge graph, a decision maker user vector is generated, including:

[0056] Vectorize the entities and relationships between entities in the decision maker's question to form a graph triple (h, r, t) covering all entities;

[0057] Among them, h represents the head entity, r represents the relationship, and t represents the tail entity;

[0058] It should be noted that a knowledge graph is a semantic network that reveals the relationships between entities;

[0059] Specifically, in step S6, if< / c> < / v> < / r> < / e> Figure 2 As shown, the decision maker user vector and the retrieval pair are input into Ripple Net iteratively, including:

[0060] Taking each of the retrieval pairs S as a node of the Ripple Net model, based on the knowledge graph G and the decision-maker user's question, the entity set of the decision-maker user in the kth round of iteration is:

[0061]

[0062] The set of triples for the kth iteration is defined as:

[0063]

[0064] When the Ripple Net model is substituted for the first iteration, the triplet of the first ring Ripple Set is obtained as follows:

[0065]

[0066] Further, based on each triple in the first ring Ripple Set, the vector r of the relationship i , vector h of the head entity i Multiply it with the vector v of the target object to obtain the head entity and relationship (h i , r i ) to obtain the decision-maker user vector representation of the first ring Ripple Set;

[0067] Optionally, the correlation score is normalized by softmax to calculate the correlation function p i as follows:

[0068]

[0069] It should be noted that p i That is, the association probability, which is the target thing v and entity h i In the relational space i The similarity in further obtains the decision maker user vector representation of this ring:

[0070]

[0071] Among them, t i ∈R d is the embedding vector of the Ripple set tail entity.

[0072] Specifically, in step S6, the decision maker user vector and the search pair are input into Ripple Net for iteration, further comprising:

[0073] When the Ripple Net model is substituted into the Ripple Net model for iteration for the kth time, the tail entity of the Ripple set in the k-1th iteration is used as the head entity of the kth iteration, thereby generating the decision-maker user vector representation of the Ripple set in the kth iteration;

[0074] Furthermore, the decision maker user vector representations of the Rippleset generated in each iteration are added together to obtain the overall decision maker user vector representation:

[0075]

[0076] Preferably, after step S6, the method further comprises:

[0077] The decision maker user retrieval recommendation vector is input into a preset knowledge graph retrieval engine, and the entity name corresponding to the decision maker user retrieval recommendation vector is output, so as to accurately provide the decision maker with recommended retrieval results based on multiple Ripplenet iterative calculations.

[0078] on the other hand, Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor (processor) 310, a communication interface (CommunicationsInterface) 320, a memory (memory) 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logic instructions in the memory 330 to execute the information retrieval method based on the knowledge graph, which includes: S1, constructing a decision maker question information vocabulary according to different element attributes for concepts, entities and relationships in the decision maker's question; S2, dividing the decision maker's question into element names and element values ​​corresponding to each element type according to the semantic algorithm, and forming element pairs with element names and corresponding element values; S3, extracting element values ​​from multiple element pairs in sequence according to the sentence order of the input decision maker's question, and then constructing a question element pattern; S4, obtaining the decision maker user identifier, obtaining each retrieval element name extracted after preprocessing the decision maker's question, obtaining the entity name of each retrieval result, and constructing a retrieval pair with the decision maker user identifier, each retrieval element name, and the entity name of each retrieval result; and storing the retrieval pair in a retrieval record library; S5, generating a decision maker user vector based on the knowledge graph; inputting the decision maker user vector and the retrieval pair into the Ripple Net model for iteration until the Ripple The decision maker user vector representation of each link in the Net model is added to form an overall decision maker user vector representation; S6, combining the overall decision maker user vector representation and the target thing, predicting the probability of the decision maker retrieving any item, and generating a decision maker user retrieval recommendation vector.

[0079] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0080] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the information retrieval method based on the knowledge graph provided by the above methods, and the method includes: S1, constructing a decision maker question information vocabulary according to different element attributes for concepts, entities and relationships in the decision maker's question; S2, dividing the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, and using element names and corresponding elements to retrieve the information retrieval method. The corresponding element values ​​constitute element pairs; S3, extract element values ​​from multiple element pairs in sequence according to the sentence order of the decision maker's question input, and then construct a question element pattern; S4, obtain the decision maker user ID, obtain each search element name extracted after preprocessing the decision maker's question, obtain the entity name of each search result, and construct a search pair with the decision maker user ID, each search element name, and the entity name of each search result; and store the search pair in the search record library; S5, based on the knowledge graph, generate a decision maker user vector; input the decision maker user vector and the search pair into the Ripple Net model for iteration until the decision maker user vector representation of each link in the Ripple Net model is added to form an overall decision maker user vector representation; S6, combined with the overall decision maker user vector representation and the target thing, predict the probability of the decision maker retrieving any item, and generate a decision maker user search recommendation vector.

[0081] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned knowledge graph-based information retrieval method.

[0082] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0083] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. 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.< / t>

Claims

1. An information retrieval method based on knowledge graph, characterized in that: include: S1. Construct the decision maker's question information word library according to different element attributes based on the concepts, entities and relationships in the decision maker's question; S2. Segment the decision maker's question into element names and element values ​​corresponding to each element type according to a semantic algorithm, and form element pairs with element names and corresponding element values; S3, extracting element values ​​from the plurality of element pairs in sequence according to the sentence order of the input decision maker's question, and then constructing a question element model; S4, obtaining a decision maker user ID, obtaining the name of each search element extracted after preprocessing the decision maker's question, obtaining the entity name of each search result, constructing a search pair with the decision maker user ID, each search element name, and the entity name of each search result; and storing the search pair in a search record library; S5. Generate a decision maker user vector based on the knowledge graph; iterate the decision maker user vector and the search pair into the RippleNet model until the decision maker user vector representation of each link of the Ripple Net model is added to form an overall decision maker user vector representation; S6, combining the overall decision maker user vector representation and the target thing, predicting the probability of the decision maker retrieving any item, and generating a decision maker user retrieval recommendation vector; Step S4 includes: The decision maker user ID U is retrieved, and the name of each search factor extracted after the decision maker's question preprocessing is F i , get the entity name of each search result as E j ; The retrieval pair S is constructed as: ; In step S5, based on the knowledge graph, a decision maker user vector is generated, including: Vectorize the entities and relationships between entities in the decision maker’s question to form a graph triple (h, r, t) covering all entities; Among them, h represents the head entity, r represents the relationship, and t represents the tail entity; The decision maker user vector and the retrieval pair are input into Ripple Net iteratively, including: Taking each of the retrieval pairs S as a node of the Ripple Net model, based on the knowledge graph G and the decision-maker user's question, the entity set of the decision-maker user in the kth round of iteration is: ; The set of triples for the kth iteration is defined as: ; When the Ripple Net model is substituted into the model for the first iteration, the triplet of the first-ring Ripple Set is obtained as follows: ; Based on each triple in the first ring Ripple Set, the vector R of the relationship i , vector h of the head entity i Multiply it with the vector v of the target thing to obtain the correlation relationship between the target thing v and the head entity and relationship in each triplet of the first iteration, thereby obtaining the decision maker user vector representation of the first-ring Ripple Set.

2. According to claim 1, the information retrieval method based on knowledge graph is characterized in that: In step S1, different element attributes include entity E, relationship R, attribute value V, concept C, space A, and time T respectively.

3. According to the information retrieval method based on knowledge graph in claim 1, it is characterized in that: In step S6, the decision maker user vector and the search pair are input into Ripple Net for iteration, further comprising: When the Ripple Net model is substituted into the Ripple Net model for iteration for the kth time, the tail entity of the Ripple set in the k-1th iteration is used as the head entity of the kth iteration, thereby generating the user representation of the Ripple set in the kth iteration.

4. The information retrieval method based on knowledge graph according to claim 1, characterized in that: After step S6, the method further includes: The decision maker user retrieval recommendation vector is input into a preset knowledge graph retrieval engine, and the entity name corresponding to the decision maker user retrieval recommendation vector is output.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the information retrieval method according to any one of claims 1 to 4 are implemented.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the information retrieval method according to any one of claims 1 to 4 are implemented.