A joint knowledge graph heat ranking method and device

Abstract

The present application proposes a joint knowledge graph heat ranking method and device, which relates to the field of information processing technology, wherein the method includes: obtaining a preset number of paper documents, performing entity recognition on the paper documents, and extracting target entities from the paper documents; extracting relationship attributes and entity attributes from the target entity, and constructing a knowledge graph based on the relationship attributes and entity attributes of the target entity; obtaining a comment text containing the target entity, and constructing an entity link with the knowledge graph based on the entity relationship contained in the comment text; determining the entity heat value based on the word frequency-inverse text frequency index value of the linked target entity and the comment text, and updating the heat attribute of the target entity in the knowledge graph based on the entity heat value. The present invention constructs a knowledge graph by extracting entities in paper documents, and predicts the heat attribute values ​​of entities in papers based on paper comments, and scientifically and reasonably gives entity heat rankings to correctly and accurately implement entity content recommendations, and can update the heat attribute values ​​of entities in real time.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a method and device for ranking the heat of a joint knowledge graph. Background Art

[0002] Knowledge graphs describe concepts, entities and their relationships in the objective world in a structured form, expressing Internet information in a form closer to the human cognitive world, providing a way to better organize, manage and understand the massive amount of Internet information. Knowledge graphs have brought vitality to Internet semantic search, and have also shown great power in intelligent question-answering, becoming the infrastructure for Internet knowledge-driven intelligent applications.

[0003] In the era of big data, the analysis of massive comment texts and public opinion texts based on knowledge graphs is particularly important. However, the application of knowledge graphs in public opinion analysis and content recommendation in the system has become a difficult problem. In related technologies, the popularity ranking of most graph entities is based on the number of text appearances and discussion frequencies as the sorting criteria, which results in insufficient text semantic analysis and inaccurate entity content recommendations. This has caused users to be dissatisfied with the recommended entity content in the domain graph system. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To achieve the above aspects, the first aspect of the present application proposes a joint knowledge graph heat ranking method, including:

[0006] Obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents;

[0007] Extracting relationship attributes and entity attributes from the target entity, and constructing a knowledge graph based on the relationship attributes and entity attributes of the target entity;

[0008] Obtaining a review text containing the target entity, and constructing an entity link with the knowledge graph according to the entity relationship contained in the review text;

[0009] An entity heat value is determined according to the word frequency-inverse text frequency index value of the linked target entity and the comment text, and a heat attribute of the target entity in the knowledge graph is updated according to the entity heat value.

[0010] Optionally, performing entity recognition on the paper document to extract target entities from the paper document includes:

[0011] Extracting document content from the paper document and converting the document content into document content in a preset format;

[0012] Sentence processing is performed on the document content in the preset format to obtain content segments corresponding to the document content, and content in the content segments is annotated to obtain annotated text corresponding to the content segments;

[0013] A training set is generated according to the annotated text, a first model is trained according to the training set, and the target entity is obtained according to a first prediction result of the first model after training.

[0014] Optionally, extracting relationship attributes and entity attributes from the target entity and constructing a knowledge graph according to the relationship attributes and entity attributes of the target entity includes:

[0015] Training a second model according to the triples of the target entity, and acquiring the relationship attribute according to a second prediction result of the trained second model;

[0016] Extracting entity attribute values ​​and corpus text of the target entity, inputting the entity attribute values ​​and corpus text into a classifier to obtain the entity attributes;

[0017] The target entity, the relationship attributes and the entity attributes are organized into tuple form and stored in a graph database to construct a knowledge graph.

[0018] Optionally, obtaining the comment text containing the target entity and constructing an entity link with the knowledge graph according to the entity relationship contained in the comment text includes:

[0019] The entity dictionary and the comment text containing the target entity are converted into vectors, and the similarity of the candidate entities is obtained, as follows:

[0020]

[0021] Among them, x i is the vector converted from the entity dictionary, y i A vector converted from the comment text containing the target entity;

[0022] According to the rule of comprehensively multiplying the mention ratio of the candidate entities and the similarity of the candidate entities, the candidate entities are sorted, and the highest-ranked candidate entity is selected to construct an entity link with the knowledge graph in the graph database.

[0023] Optionally, before the word frequency-inverse text frequency index value of the target entity linked to the comment text, the method further includes:

[0024] Calculate the word frequency TF value, that is, the frequency of the target entity appearing in the comment text, as follows:

[0025]

[0026] Among them, n i,j is the target entity in file d j The number of times it appears in k n k,j For file d j The sum of the number of occurrences of all words in ;

[0027] Calculate the inverse text frequency IDF value, the formula is as follows:

[0028]

[0029] Where |D| is the total number of documents in the corpus, |{j:t i ∈d j}| indicates that the target entity t is included i The number of files;

[0030] Calculate the term frequency-inverse text frequency index value TF-IDF value, the formula is as follows:

[0031] TF-IDF=TF*IDF.

[0032] Optionally, determining the entity heat value includes:

[0033] Obtain the target entity click volume and user comment volume, and obtain the target entity's own popularity value based on the entity click volume and user comment volume;

[0034] Obtain the text comment time according to the comment text interface, and obtain the update time according to the text comment time;

[0035] According to the TF-IDF value, the heat value of the target entity itself and the update time, the heat value of the entity is obtained, and the formula is as follows:

[0036]

[0037] Among them, H is the entity heat value, W is the heat value of the target entity itself, T is the update time, and G is the gravity weight index.

[0038] Optionally, the updating of the heat attribute of the target entity in the knowledge graph according to the entity heat value further includes:

[0039] The entity heat value is updated to the graph database, and the heat value attribute of the node is obtained to update the comment text data and the entity heat value.

[0040] The second aspect of the present application proposes a heat ranking device for a joint knowledge graph, comprising:

[0041] An entity extraction module is used to obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents;

[0042] A knowledge graph construction module, used to extract relationship attributes and entity attributes from the target entity, and construct a knowledge graph based on the relationship attributes and entity attributes of the target entity;

[0043] A link building module, used to obtain a comment text containing the target entity, and build an entity link with the knowledge graph according to the entity relationship contained in the comment text;

[0044] A heat value determination module is used to determine the entity heat value according to the word frequency-inverse text frequency index value of the linked target entity and the comment text, and update the heat attribute of the target entity in the knowledge graph according to the entity heat value.

[0045] The third aspect of the present application provides a data processing electronic device, comprising:

[0046] processor;

[0047] a memory for storing instructions executable by the processor;

[0048] The processor is configured to execute the instructions to implement any one of the methods described in the first aspect above.

[0049] The fourth aspect of the present application proposes a computer-readable storage medium. When the instructions in the storage medium are executed by a processor of a data processing electronic device, the data processing electronic device can execute any of the methods described in the first aspect.

[0050] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0051] By extracting entities from paper documents to build a knowledge graph, and predicting the hot attribute values ​​of entities in the paper based on paper comments, a scientific and reasonable entity hotness ranking is given to correctly and accurately implement entity content recommendations, and the hot attribute values ​​of entities can be updated in real time.

[0052] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0054] Figure 1 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0055] Figure 2 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0056] Figure 3 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0057] Figure 4 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0058] Figure 5 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0059] Figure 6 is a flowchart of a heat ranking method of a joint knowledge graph according to an exemplary embodiment;

[0060] Figure 7 is a block diagram of a heat ranking device for a joint knowledge graph according to an exemplary embodiment;

[0061] Figure 8 is a schematic block diagram of an electronic device. DETAILED DESCRIPTION

[0062] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0063] The following describes a method and device for hotness ranking of a joint knowledge graph in an embodiment of the present application with reference to the accompanying drawings.

[0064] Figure 1 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 1 As shown, the method comprises the following steps:

[0065] Step 101, obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents.

[0066] In this embodiment, a certain number of paper documents are prepared, and entities in the document content are extracted according to predefined entity types, where the entity type can be attributes such as names of people, institutions, and places.

[0067] In a possible embodiment, the preset number of paper documents is 3000, the predefined entity is "praying mantis", and the part containing "praying mantis" in the document content is identified and extracted.

[0068] Step 102, extracting relationship attributes and entity attributes from the target entity, and constructing a knowledge graph based on the relationship attributes and entity attributes of the target entity.

[0069] In this embodiment, relationship attributes and entity attributes in the target entity are extracted, wherein the relationship attributes are used to describe the relationship and connection between entities, and the entity attributes are used to describe the content and characteristics of the entity.

[0070] In a possible embodiment, "student" uses "room card", "student" and "meal card" are different entities, and "like to eat" is the relationship between the two entities.

[0071] In a possible embodiment, the entity "student" can be described by attributes such as "name", "class", and "department".

[0072] Step 103, obtaining the comment text containing the target entity, and constructing an entity link with the knowledge graph according to the entity relationship contained in the comment text.

[0073] In this embodiment, the comment text is the public opinion text. The graph is generally applied on the system, and users will discuss the entity information. The content of the discussion is the comment text. Sometimes the discussed entity or similar entities will appear in the discussion text. The comment text containing the entity or similar entity will be linked to the knowledge graph.

[0074] Step 104, determining an entity heat value according to the word frequency-inverse text frequency index value of the linked target entity and the comment text, and updating the heat attribute of the target entity in the knowledge graph according to the entity heat value.

[0075] In this embodiment, the TF-IDF value of the entity for establishing entity link in the comment text is calculated, and the word frequency value and the inverse text frequency index value are calculated in turn to determine the probability of the entity appearing in the comment text and the probability of the file containing the entity appearing in the total corpus, so as to evaluate the popularity of the entity.

[0076] In a possible embodiment, "canary" appears 20 times in the comment text, the entities that appear in the comment text are 10,000 times (repeated entities can be repeatedly overlapped), there are 1,000 files containing "canary", and the total number of files in the total corpus is 10,000,000.

[0077] In this embodiment, a knowledge graph is constructed by extracting entities from paper documents, and the popularity attribute values ​​of entities in the paper are predicted based on paper comments. The entity popularity ranking is given scientifically and reasonably to correctly and accurately implement entity content recommendations, and the popularity attribute values ​​of entities can be updated in real time.

[0078] Figure 2 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 2 As shown, step 101 also includes:

[0079] Step 201, extracting document content from the paper document and converting the document content into document content in a preset format.

[0080] In this embodiment, the paper itself is in PDF format, and PDFplumber technology can be used to extract page text, tables and other contents of the PDF file, and convert the PDF file into corresponding txt text, wherein PDFplumber is a tool for converting PDF format.

[0081] Step 202, sentence processing is performed on the document content in the preset format to obtain content segments corresponding to the document content, and the content in the content segments is annotated to obtain annotated text corresponding to the content segments.

[0082] In a possible embodiment, the converted paper txt text is divided into an abstract and a main text and all converted into a half-width format, and regular expressions are used to separate sentences according to punctuation marks and line breaks, and the text content of the divided sentences is annotated with a BMES sequence, annotated as a text method of single-word corresponding labels, where the BMES sequence annotation is explained as: B represents the first value of a word, M represents the middle position of a word, E represents the end position of a word, and S represents a single word.

[0083] Step 203: Generate a training set according to the annotated text, train a first model according to the training set, and obtain the target entity according to a first prediction result of the first model after training.

[0084] In this embodiment, a first model is trained based on the training set text, and prediction results are obtained based on the trained model, and the prediction results are converted into entities with corresponding labels.

[0085] In one possible embodiment, the process of training the first model is: input the labeled text, and convert it into a vocabulary ID through BertTokenizer, randomly replace 15% of the IDs with the ID of the word [mask], connect the CRF layer to the Bert model output, and train the model until the model converges.

[0086] This embodiment organizes the document content by converting the format, sentence division, annotation, etc. of the paper, and obtains the target entity by training the model, so as to obtain the entity more conveniently and quickly.

[0087] Figure 3 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 3 As shown, step 102 also includes:

[0088] Step 301: train a second model according to the triples of the target entity, and obtain the relationship attribute according to a second prediction result of the trained second model.

[0089] In this embodiment, the triples are trained and the relationship attributes between entities are obtained according to the second model. The triples are generally composed of "<entity>+relationship+<entity>".

[0090] In a possible implementation example, the second model is a BERT+CRF model.

[0091] Step 302, extracting entity attribute values ​​and corpus text of the target entity, inputting the entity attribute values ​​and corpus text into a classifier, and obtaining the entity attributes;

[0092] In this embodiment, meta-pattern candidates can be generated based on MetaPAD mining, quality assessment functions of candidate patterns can be learned, and feature selection has rich domain-independent context features.

[0093] A pair of sub-schema entities <mp l ,mp r The expected frequency formula of > is:

[0094] μ0(c( <mp l ,mp r >))=L·p(mp l )·p(mp r )

[0095] in is the empirical probability of the pattern. Use the Z index to measure the probability of a pair of sub-pattern entities. <mp l ,mp r > constitutes the best index in the corpus, a high Z-index indicates that the pattern acts as a complete semantic unit in the context.

[0096]

[0097] According to the above Z values, multiple meta-pattern groups can be formed to extract aggregation relationships from the corpus text, and then an attribute classifier can be trained.

[0098] In a possible embodiment, the attribute classifier may be a random forest-based attribute classifier.

[0099] Input entities and text into the above classifier to obtain their entity attributes, and bind the attributes to the corresponding entities through neo4j. Among them, since the knowledge graph contains entities, attributes, relationships, etc., common relational databases such as MySQL cannot well reflect these characteristics of the data. Therefore, the storage of knowledge graph data is generally done using graph databases, and neo4j is the most common graph database.

[0100] In a possible embodiment, binding attributes to corresponding entities may be: binding attributes such as "name", "class", and "department" to the "student" entity.

[0101] Step 302: organize the target entity, the relationship attributes, and the entity attributes into tuple form and store them in a graph database to construct a knowledge graph.

[0102] In one possible embodiment, Neo4j is used to build a knowledge graph based on Py2neo, which is a client library and toolkit that can use Neo4j within Python applications and command lines.

[0103] This embodiment obtains the relationship attributes and entity attributes of the entity, and binds the attributes through a graph database to create a knowledge graph, which facilitates the construction of the knowledge graph in the system.

[0104] Figure 4 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 4 As shown, step 103 also includes:

[0105] Step 401, converting the entity dictionary and the comment text containing the target entity into vectors, and obtaining the similarity of the candidate entities, is formulated as follows:

[0106]

[0107] Among them, x i is the vector converted from the entity dictionary, y i is the converted vector of the comment text containing the target entity.

[0108] In this embodiment, the comment text under the knowledge graph is obtained according to the graph system, and then the entity dictionary and the comment text string similarity algorithm that mentions the entity are used to embed the two texts into vectors, and then the cosine similarity formula is used to obtain the similarity of the entity. If the similarity is high, it is considered that the entity fuzzy match is successful, and a candidate entity with high similarity is generated. The property of the embedding vector is that objects corresponding to vectors with similar distances have similar meanings.

[0109] Step 402, sorting the candidate entities according to the rule of comprehensive multiplication of the mention ratio of the candidate entities and the similarity of the candidate entities, selecting the highest-ranked candidate entity and constructing an entity link with the knowledge graph in the graph database.

[0110] In this embodiment, the candidate entities are sorted according to preset rules, and the highest-ranked candidate entity is selected to establish a link with the knowledge graph in the graph database.

[0111] This embodiment calculates and sorts the similarities of candidate entities in the comment text to establish a link between the comment text and the knowledge graph, so as to facilitate the subsequent acquisition of entity heat values.

[0112] Figure 5 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 5 As shown, including:

[0113] Step 501, calculate the word frequency TF value, that is, the frequency of the target entity appearing in the comment text, which is formulated as follows:

[0114]

[0115] Among them, n i,j is the target entity in file d j The number of times it appears in k n k,j For file d j The sum of the number of occurrences of all words in .

[0116] Step 502, calculate the inverse text frequency IDF value, which is formulated as follows:

[0117]

[0118] Where |D| is the total number of documents in the corpus, |{j:t i ∈d j}| indicates that the target entity t is included i The number of files;

[0119] Step 503, calculate the term frequency-inverse text frequency index value TF-IDF value, which is formulated as follows:

[0120] TF-IDF=TF*IDF.

[0121] In this embodiment, the term frequency TF value and the inverse text frequency IDF value are calculated to obtain the TF-IDF value.

[0122] Figure 6 is a flowchart of a heat ranking method for a joint knowledge graph according to an exemplary embodiment. Figure 6 As shown, step 104 also includes:

[0123] Step 601, obtaining the target entity click volume and user comment volume, and obtaining the target entity's own heat value according to the entity click volume and user comment volume.

[0124] In this embodiment, the number of clicks and user comments of the target entity is obtained according to the graph system, and the heat value of the target entity itself is obtained according to the preset weighting rules.

[0125] In a possible embodiment, the weights of entity click volume and user comment volume are 0.01 and 0.99 respectively.

[0126] Step 602, obtaining the text comment time according to the comment text interface, and obtaining the update time according to the text comment time.

[0127] In this embodiment, the text comment update time is obtained according to the system comment interface.

[0128] Step 603, according to the TF-IDF value, the heat value of the target entity itself and the update time, the entity heat value is obtained, which is formulated as follows:

[0129]

[0130] Among them, H is the entity heat value, W is the heat value of the target entity itself, T is the update time, and G is the gravity weight index.

[0131] Optionally, it also includes;

[0132] The entity heat value is updated to the graph database, and the heat value attribute of the node is obtained to update the comment text data and the entity heat value.

[0133] In a possible embodiment, based on the calculated heat value H, py2neo is used to update it to the graph database as the heat value attribute of the node, and the schedule module in python is used to regularly update the comment text data and heat value.

[0134] This embodiment predicts the heat attribute values ​​of entities in a paper based on paper reviews, and gives a scientific and reasonable heat ranking of the entities to correctly and accurately implement entity content recommendations, and can update the heat attribute values ​​of the entities in real time.

[0135] Figure 7 is a block diagram 700 of a heat ranking device for a joint knowledge graph according to an exemplary embodiment, referring to Figure 7 The device includes: an entity extraction module 701, a knowledge graph construction module 702, a link construction module 703 and a heat value determination module 704.

[0136] The entity extraction module 710 is used to obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents;

[0137] A knowledge graph construction module 720, used to extract relationship attributes and entity attributes from the target entity, and construct a knowledge graph according to the relationship attributes and entity attributes of the target entity;

[0138] A link building module 730 is used to obtain a comment text containing the target entity, and build an entity link with the knowledge graph according to the entity relationship contained in the comment text;

[0139] The heat value determination module 740 is used to determine the entity heat value according to the word frequency-inverse text frequency index value of the linked target entity and the comment text, and update the heat attribute of the target entity in the knowledge graph according to the entity heat value.

[0140] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0141] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0142] like Figure 8As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0143] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0144] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as the voice instruction response method. For example, in some embodiments, the voice instruction response method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the voice instruction response method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the voice instruction response method in any other appropriate manner (e.g., by means of firmware).

[0145] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0146] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0147] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0148] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0149] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0150] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0151] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0152] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A heat ranking method for joint knowledge graphs, characterized in that: include: Obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents; Extracting relationship attributes and entity attributes from the target entity, and constructing a knowledge graph based on the relationship attributes and entity attributes of the target entity; Obtaining a review text containing the target entity, and constructing an entity link with the knowledge graph according to the entity relationship contained in the review text; Calculate the word frequency-inverse text frequency index value of the comment text, determine the entity heat value according to the linked target entity and the word frequency-inverse text frequency index value of the comment text, and update the heat attribute of the target entity in the knowledge graph according to the entity heat value; Determining the entity heat value includes: Obtain the target entity's click volume and user comment volume, and obtain the target entity's own popularity value based on the entity's click volume and user comment volume; Obtain the text comment time according to the comment text interface, and obtain the update time according to the text comment time; According to the word frequency-inverse text frequency index value of the comment text, the heat value of the target entity itself and the update time, the entity heat value is obtained, and the formula is as follows: Wherein, H is the entity heat value, W is the target entity heat value, T is the update time, G is the gravity weight index, It is the word frequency-inverse text frequency index value of the comment text.

2. The method according to claim 1, characterized in that The performing entity recognition on the paper document and extracting the target entity from the paper document includes: Extracting document content from the paper document and converting the document content into document content in a preset format; Sentence processing is performed on the document content in the preset format to obtain content segments corresponding to the document content, and content in the content segments is annotated to obtain annotated text corresponding to the content segments; A training set is generated according to the annotated text, a first model is trained according to the training set, and the target entity is obtained according to a first prediction result of the first model after training.

3. The method according to claim 2, characterized in that The step of extracting relationship attributes and entity attributes from the target entity and constructing a knowledge graph according to the relationship attributes and entity attributes of the target entity includes: Training a second model according to the triples of the target entity, and acquiring the relationship attribute according to a second prediction result of the trained second model; Extracting entity attribute values ​​and corpus text of the target entity, inputting the entity attribute values ​​and corpus text into a classifier to obtain the entity attributes; The target entity, the relationship attributes and the entity attributes are organized into tuple form and stored in a graph database to construct a knowledge graph.

4. The method according to claim 3, characterized in that: The obtaining of the comment text containing the target entity and constructing an entity link with the knowledge graph according to the entity relationship contained in the comment text includes: The entity dictionary and the comment text containing the target entity are converted into vectors, and the similarity of the candidate entities is obtained, as follows: ； in, is the vector converted from the entity dictionary, A vector converted from the comment text containing the target entity; According to the rule of comprehensively multiplying the mention ratio of the candidate entities and the similarity of the candidate entities, the candidate entities are sorted, and the highest-ranked candidate entity is selected to construct an entity link with the knowledge graph in the graph database.

5. The method according to claim 4, characterized in that Before the word frequency-inverse text frequency index value of the target entity linked to the comment text, it also includes: Calculate the word frequency TF value, that is, the frequency of the target entity appearing in the comment text, as follows: ； in, is the target entity in the file The number of times it appears in For File The sum of the number of occurrences of all words in ; Calculate the inverse text frequency IDF value, the formula is as follows: ； Where |D| is the total number of documents in the corpus, Indicates that the target entity is included The number of files; Calculate the term frequency-inverse text frequency index value TF-IDF value, the formula is as follows: 。 6. The method according to claim 5, characterized in that The updating of the heat attribute of the target entity in the knowledge graph according to the entity heat value also includes: The entity heat value is updated to the graph database, and the heat value attribute of the node is obtained to update the comment text data and the entity heat value.

7. A heat ranking device for a joint knowledge graph, characterized in that: include: An entity extraction module is used to obtain a preset number of paper documents, perform entity recognition on the paper documents, and extract target entities from the paper documents; A knowledge graph construction module, used to extract relationship attributes and entity attributes from the target entity, and construct a knowledge graph based on the relationship attributes and entity attributes of the target entity; A link building module, used to obtain a comment text containing the target entity, and build an entity link with the knowledge graph according to the entity relationship contained in the comment text; A heat value determination module, used to calculate the word frequency-inverse text frequency index value of the comment text, determine the entity heat value according to the linked target entity and the word frequency-inverse text frequency index value of the comment text, and update the heat attribute of the target entity in the knowledge graph according to the entity heat value; Determining the entity heat value includes: Obtain the target entity's click volume and user comment volume, and obtain the target entity's own popularity value based on the entity's click volume and user comment volume; Obtain the text comment time according to the comment text interface, and obtain the update time according to the text comment time; According to the word frequency-inverse text frequency index value of the comment text, the heat value of the target entity itself and the update time, the entity heat value is obtained, and the formula is as follows: Wherein, H is the entity heat value, W is the target entity heat value, T is the update time, G is the gravity weight index, It is the word frequency-inverse text frequency index value of the comment text.

8. A data processing electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, when instructions in the storage medium are executed by a processor of a data processing electronic device, the data processing electronic device is enabled to perform the method according to any one of claims 1 to 7.

Images