Entity linking method and apparatus, electronic device, and storage medium

Abstract

The disclosure provides an entity linking method and device, electronic equipment and storage medium, and relates to the fields of knowledge graph, deep learning and natural language processing and the like artificial intelligence. The method can include: performing entity recognition on first text information corresponding to a first entity to be processed, the first text information including introduction information of the first entity; for any identified entity, finding an entity corresponding to the entity from entities recorded in a knowledge base, and taking the found entity as a recall entity; obtaining second text information corresponding to each recall entity, the second text information including introduction information of the recall entity; and selecting an entity corresponding to the first entity from each recall entity according to the first text information and the second text information. The scheme disclosed in the disclosure can save labor and time costs.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, in particular to an entity linking method and device, an electronic device and a storage medium in the fields of knowledge graph, deep learning and natural language processing. BACKGROUND

[0002] Entity Linking refers to a task of linking a certain entity, such as an entity in a text, to an entity in a knowledge graph / knowledge base. In many scenarios, there is a need for entity linking. For example, for a certain account a (corresponding to an entity of a person type) on a certain content creation platform, the account a can be linked to a corresponding entity b in a knowledge base, so that a user can enter the author homepage of the account a by clicking a relevant interface area while browsing the structured information of the entity b in the knowledge base, and view the real-time dynamic information published by the author.

[0003] Currently, an entity linking model trained is usually used to implement the entity linking. In order to obtain a high-performance entity linking model, a large amount of training corpus needs to be collected or labeled manually, thereby a large amount of human and time costs are consumed. SUMMARY

[0004] The present disclosure provides an entity linking method, device, electronic device and storage medium.

[0005] An entity linking method comprises:

[0006] performing entity recognition on first text information corresponding to a first entity to be processed, the first text information including introduction information about the first entity;

[0007] for any recognized entity, finding an entity corresponding to the entity from entities recorded in a knowledge base, and taking the found entity as a recall entity;

[0008] obtaining second text information corresponding to each recall entity, the second text information including introduction information about the recall entity;

[0009] selecting an entity corresponding to the first entity from the recall entities according to the first text information and the second text information.

[0010] An entity linking device comprises an identification module, a recall module, an obtaining module and a linking module.

[0011] The identification module is configured to perform entity recognition on first text information corresponding to a first entity to be processed, the first text information including introduction information about the first entity;

[0012] The recall module is configured to, for any identified entity, find an entity corresponding to the identified entity from entities recorded in the knowledge base, and take the found entity as a recall entity.

[0013] The acquisition module is configured to acquire second text information corresponding to each recall entity, wherein the second text information includes introduction information about the recall entity.

[0014] The linking module is configured to select an entity corresponding to the first entity from the recall entities according to the first text information and the second text information.

[0015] An electronic device includes:

[0016] at least one processor; and

[0017] a memory communicatively connected to the at least one processor; wherein

[0018] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0019] A non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method as described above.

[0020] A computer program product including computer programs / instructions, which, when executed by a processor, implement the method as described above.

[0021] One embodiment in the above disclosure has the following advantages or beneficial effects: for a first entity to be processed, a corresponding entity in the knowledge base can be automatically determined through a series of processes, i.e., entity linking is achieved, without using an entity linking model, thereby saving manpower and time costs and ensuring the accuracy of the processing result.

[0022] It should be understood that the content described in this part is not intended to identify key or important features of embodiments of the disclosure, nor is it used to limit the scope of the disclosure. Other features of the disclosure will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0023] The accompanying drawings serve to better understand the present scheme and do not limit the disclosure. Among them:

[0024] Figure 1 Flowchart of the first embodiment of the entity linking method according to the disclosure;

[0025] Figure 2This is a flowchart of a second embodiment of the entity linking method described in this disclosure;

[0026] Figure 3 This is a schematic diagram of the composition structure of embodiment 300 of the entity linking device described in this disclosure;

[0027] Figure 4 A schematic block diagram of an electronic device 400 that can be used to implement embodiments of the present disclosure is shown. Detailed Implementation

[0028] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0029] Furthermore, it should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0030] Figure 1 This is a flowchart of a first embodiment of the entity linking method described in this disclosure. Figure 1 As shown, the specific implementation methods are as follows.

[0031] In step 101, entity recognition is performed on the first text information corresponding to the first entity to be processed. The first text information includes introductory information about the first entity.

[0032] In step 102, for any identified entity, the entity corresponding to that entity is retrieved from the entities recorded in the knowledge base, and the retrieved entity is used as the recall entity.

[0033] In step 103, the second text information corresponding to each recalled entity is obtained, and the second text information includes introductory information about the recalled entity.

[0034] In step 104, the entity corresponding to the first entity is selected from each recalled entity based on the first text information and the second text information.

[0035] It can be seen that in the scheme of the above method embodiment, for the first entity to be processed, the corresponding entity in the knowledge base can be automatically determined through a series of processes, that is, entity linking is realized, without using an entity linking model, thereby saving manpower and time costs and ensuring the accuracy of the processing results.

[0036] For the first entity to be processed, entity recognition can be performed on the first text information corresponding to the first entity. Preferably, the first text information can be subjected to a predetermined type of entity recognition, which is the same type as that of the first entity.

[0037] The predetermined type is not limited in type, for example, it can be a person type, that is, a person type entity can be identified from the first text information through entity recognition technology. For example, the first entity can refer to an account on a content creation platform, and the first text information corresponding to the first entity can include introduction information about the first entity, such as introduction information about the author corresponding to the account.

[0038] For any entity identified, the corresponding entity can be found from the entities recorded in the knowledge base. Preferably, entities that do not meet the requirements can be filtered out from the identified entities first, and then for any entity remaining after filtering, the corresponding entity can be found from the entities recorded in the knowledge base.

[0039] How to filter out entities that do not meet the requirements from the identified entities is not limited. For example, the identified entities can be sorted in descending order of relevance to the first entity, and the top 2 or top 3 entities after sorting can be retained, and the others can be filtered out, that is, the entities that do not meet the requirements. Among them, a pre-trained evaluation model can be used to determine the relevance between each entity and the first entity.

[0040] For example, the first entity is an account on a content creation platform, and the account is "Zhang San". Assuming that 4 entities are identified from the first text information, namely "Zhang San", "Li Si", "Wang Wu" and "Zhao Liu", and the sorted order is "Zhang San", "Wang Wu", "Li Si" and "Zhao Liu", then "Li Si" and "Zhao Liu" can be filtered out, and the remaining entities "Zhang San" and "Wang Wu" are obtained.

[0041] Through the above processing, interference entities with low relevance can be filtered out, thereby reducing the workload of subsequent processing and improving processing efficiency.

[0042] For any entity remaining after filtering, the corresponding entity can be found from the entities recorded in the knowledge base, and the found entity can be used as a recall entity.

[0043] For example, the remaining entities after filtering are "Zhang San" and "Wang Wu", for the entity "Zhang San", the corresponding entity can refer to the entity recorded in the knowledge base with the name "Zhang San", similarly, for the entity "Wang Wu", the corresponding entity can refer to the entity recorded in the knowledge base with the name "Wang Wu", in addition, the same name may correspond to multiple different entities for the knowledge base, assuming that there are 3 entities with the name "Zhang San" and 2 entities with the name "Wang Wu", then according to "Zhang San" and "Wang Wu", 5 recall entities can be obtained.

[0044] If only the first text information is subjected to the predetermined type of entity recognition, then preferably, entities that are not of the predetermined type can also be filtered out from each recall entity. For example, if the predetermined type is a person type, then entities that are not of the person type can be filtered out from each recall entity. Through the filtering process, the workload of subsequent processing can be further reduced and the processing efficiency can be improved, etc.

[0045] After that, the second text information corresponding to each of the remaining recall entities after filtering can be obtained respectively, and the second text information includes introduction information of the recall entity. For example, the second text information corresponding to each of the remaining recall entities after filtering can be obtained from the knowledge base respectively.

[0046] Further, according to the first text information and the second text information, the entity corresponding to the first entity can be selected from each of the recall entities. Preferably, the relevance score between each second text information and the first text information can be obtained respectively, and the entity corresponding to the first entity can be selected from each of the recall entities according to the relevance score.

[0047] Among them, the way of obtaining the relevance score between each second text information and the first text information can include the following ways 1), 2) and 3).

[0048] Way 1): keyword extraction is performed on the first text information, for any second text information, the keywords appearing in the second text information are selected from the extracted keywords respectively, and the relevance score between the second text information and the first text information is determined according to the selection result.

[0049] Way 2): the first text information is subjected to predetermined type keyword extraction, and the extracted keywords are taken as the core keywords of the first text information, for any second text information, the predetermined type keyword extraction is performed on the second text information respectively, and the extracted keywords are taken as the core keywords of the second text information, the core keywords of the second text information and the core keywords of the first text information are compared to determine the relevance score between the second text information and the first text information.

[0050] The third mode is: keyword extraction is performed on the first text information, for any second text information, keywords appearing in the second text information are screened out from the extracted keywords respectively, and a first correlation score between the second text information and the first text information is determined according to the screening result; the first text information is subjected to predetermined type keyword extraction, and the extracted keywords are taken as core keywords of the first text information, for any second text information, the second text information is subjected to the predetermined type keyword extraction respectively, and the extracted keywords are taken as core keywords of the second text information, and a second correlation score between the second text information and the first text information is determined by comparing the core keywords of the second text information and the core keywords of the first text information; for any second text information, a final correlation score between the second text information and the first text information is determined by combining the first correlation score and the second correlation score respectively.

[0051] In the first mode, the first text information can be subjected to keyword extraction by using an existing keyword extraction technology, and correspondingly, for any second text information, keywords appearing in the second text information can be screened out from the extracted keywords respectively, and a correlation score between the second text information and the first text information can be determined according to the screening result. For example, each keyword extracted can be assigned a corresponding weight, and the weights corresponding to the keywords appearing in the second text information can be added together, and the sum is taken as the correlation score between the second text information and the first text information. The correlation between each keyword extracted and the first text information can be determined by using an evaluation model trained in advance, and each keyword can be assigned a corresponding weight according to the correlation. The higher the correlation is, the greater the weight can be.

[0052] In the second mode, the first text information can be subjected to predetermined type keyword extraction, and the extracted keywords are taken as core keywords of the first text information. The core keywords usually include the keywords extracted in the first mode, and the specific types of the predetermined type can be determined according to actual needs. For example, the core keywords can include professional information, organization information, work information and other keywords reflecting the characteristics of the entity type. Correspondingly, for any second text information, the second text information can be subjected to the predetermined type keyword extraction respectively, and the extracted keywords are taken as core keywords of the second text information. Then, the correlation score between the second text information and the first text information can be determined by comparing the core keywords of the second text information and the core keywords of the first text information. For example, the difference between the core keywords of the second text information and the core keywords of the first text information can be compared, and the smaller the difference is, the higher the correlation score can be.

[0053] The core keywords are usually proper nouns. In addition, for the proper nouns that are not reasonably segmented in the second text information, corresponding merging or splitting processing can be performed.

[0054] In mode 3), for any second text information, the first relevance score between the second text information and the first text information can be determined in the manner of mode 1), and the second relevance score between the second text information and the first text information can be determined in the manner of mode 2), and then the final relevance score between the second text information and the first text information can be determined in combination with the first relevance score and the second relevance score. For example, the first relevance score and the second relevance score can be weighted and added, and the obtained result is taken as the final relevance score between the second text information and the first text information. The weights corresponding to the first relevance score and the second relevance score can be determined according to actual needs.

[0055] Preferably, when the entity corresponding to the first entity is selected from the recall entities according to the relevance scores, the relevance score with the maximum value can be first selected from the relevance scores, and then it can be determined whether the relevance score with the maximum value is greater than a predetermined threshold. If yes, the recall entity corresponding to the relevance score with the maximum value is taken as the entity corresponding to the first entity, otherwise, it is considered that the entity linking fails. The specific value of the predetermined threshold can be determined according to actual needs.

[0056] Through the above processing, the entity corresponding to the first entity can be efficiently and accurately selected from the recall entities by means of the relevance scores, and in actual application, the relevance scores between the first text information and the second text information can be obtained by arbitrarily selecting mode 1), mode 2) or mode 3) above, which is very flexible and convenient. In addition, the entity corresponding to the first entity is finally determined in combination with the maximum value and the predetermined threshold, thereby ensuring the accuracy of the entity linking result and the like.

[0057] Taking the entity of the first entity and the like as an example, Figure 2 The flowchart of the second embodiment of the entity linking method of the present disclosure is shown in FIG. 2. Figure 2 As shown in FIG. 2, the following specific implementation modes are included.

[0058] In step 201, for the first entity to be processed, the first text information corresponding thereto is obtained, the first entity is an entity of a person type, and the first text information includes introduction information about the first entity.

[0059] In step 202, the entity of a person type is identified from the first text information, and the entities that do not meet the requirements are filtered out from the identified entities.

[0060] In step 203, for any entity remaining after filtering, the corresponding entity of the entity is found from the entities recorded in the knowledge base respectively, and the found entity is taken as a recall entity, and entities that are not of the person type are filtered out from the recall entities.

[0061] In step 204, the second text information corresponding to each recall entity remaining after filtering is acquired respectively, and the second text information includes introduction information of the recall entity.

[0062] In step 205, keyword extraction is performed on the first text information, for any second text information, keywords appearing in the second text information are selected from the extracted keywords respectively, and a first relevance score between the second text information and the first text information is determined according to the selection result.

[0063] In step 206, the first text information is subjected to predetermined type keyword extraction, and the extracted keywords are taken as core keywords of the first text information, for any second text information, the second text information is subjected to the predetermined type keyword extraction, and the extracted keywords are taken as core keywords of the second text information, and a second relevance score between the second text information and the first text information is determined by comparing the core keywords of the second text information and the core keywords of the first text information.

[0064] In step 207, for any second text information, a final relevance score between the second text information and the first text information is determined by combining the first relevance score and the second relevance score respectively.

[0065] In step 208, the relevance score with the largest value is selected from the relevance scores, and if the relevance score with the largest value is greater than a predetermined threshold, the recall entity corresponding to the relevance score with the largest value is taken as the entity corresponding to the first entity.

[0066] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a combination of a series of actions, but those skilled in the art should know that the present disclosure is not limited by the order of the described actions, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily necessary for the present disclosure. In addition, the parts not described in detail in a certain embodiment can refer to the related description in other embodiments.

[0067] The above is the introduction of the method embodiment, and the scheme of the present disclosure will be further described through the device embodiment below.

[0068] Figure 3Fig. 1 shows a schematic diagram of an embodiment of the entity linking apparatus according to the present disclosure. As shown in Fig. 1, the entity linking apparatus comprises an identifying module 101, a retrieving module 102, an obtaining module 103, and a linking module 104. Figure 3

[0069] The identifying module 101 is configured to perform entity recognition on first text information corresponding to a first entity to be processed, the first text information including introduction information about the first entity.

[0070] The retrieving module 102 is configured to, for any entity recognized by the identifying module 101, find an entity corresponding to the entity from entities recorded in a knowledge base, and take the found entity as a retrieved entity.

[0071] The obtaining module 103 is configured to obtain second text information corresponding to each retrieved entity, the second text information including introduction information about the retrieved entity.

[0072] The linking module 104 is configured to select an entity corresponding to the first entity from the retrieved entities according to the first text information and the second text information.

[0073] Preferably, the identifying module 101 can further filter out entities that do not meet the requirements from the recognized entities, and correspondingly, the retrieving module 102 can, for any entity remaining after the filtering, find an entity corresponding to the entity from entities recorded in the knowledge base, and take the found entity as a retrieved entity.

[0074] How to filter out entities that do not meet the requirements from the recognized entities is not limited. For example, the recognized entities can be sorted in descending order of relevance to the first entity, and entities in the top 2 or top 3 positions after the sorting can be retained, and other entities can be filtered out, the filtered-out entities being the entities that do not meet the requirements.

[0075] Preferably, the identifying module 101 can further perform entity recognition of a predetermined type on the first text information, the predetermined type being the same as the type of the first entity. The predetermined type is not limited in type, for example, it can be a person type, i.e., a person type entity can be recognized from the first text information by the entity recognition technology.

[0076] Correspondingly, the retrieving module 102 can further filter out entities that are not of the predetermined type from the retrieved entities, for example, entities that are not of the person type can be filtered out from the retrieved entities.

[0077] Then, the obtaining module 103 can obtain second text information corresponding to each retrieved entity remaining after the filtering, the second text information including introduction information about the retrieved entity. For example, the obtaining module 103 can obtain second text information corresponding to each retrieved entity remaining after the filtering from the knowledge base. ​

[0078] Further, the linking module 304 can select the entity corresponding to the first entity from the recall entities according to the first text information and the second text information. Preferably, the linking module 304 can obtain a relevance score between each second text information and the first text information respectively, and select the entity corresponding to the first entity from the recall entities according to the relevance score.

[0079] The manner in which the linking module 304 obtains the relevance score between each second text information and the first text information respectively can include the following manners 1), 2) and 3).

[0080] Manner 1): keyword extraction is performed on the first text information, for any second text information, keywords appearing in the second text information are selected from the extracted keywords respectively, and the relevance score between the second text information and the first text information is determined according to the selection result.

[0081] Manner 2): a predetermined type of keyword extraction is performed on the first text information, and the extracted keywords are taken as core keywords of the first text information, for any second text information, the predetermined type of keyword extraction is performed on the second text information respectively, and the extracted keywords are taken as core keywords of the second text information, and the relevance score between the second text information and the first text information is determined by comparing the core keywords of the second text information and the core keywords of the first text information.

[0082] Manner 3): keyword extraction is performed on the first text information, for any second text information, keywords appearing in the second text information are selected from the extracted keywords respectively, and the first relevance score between the second text information and the first text information is determined according to the selection result; a predetermined type of keyword extraction is performed on the first text information, and the extracted keywords are taken as core keywords of the first text information, for any second text information, the predetermined type of keyword extraction is performed on the second text information respectively, and the extracted keywords are taken as core keywords of the second text information, and the second relevance score between the second text information and the first text information is determined by comparing the core keywords of the second text information and the core keywords of the first text information; for any second text information, the final relevance score between the second text information and the first text information is determined by combining the first relevance score and the second relevance score respectively.

[0083] Preferably, when selecting the entity corresponding to the first entity from the recall entities according to the correlation scores, the linking module 304 can first select the correlation score with the maximum value from the correlation scores, and further determine whether the correlation score with the maximum value is greater than a predetermined threshold. If yes, the recall entity corresponding to the correlation score with the maximum value is taken as the entity corresponding to the first entity, otherwise, it is considered that the entity linking fails this time.

[0084] Figure 3 The specific working process of the device embodiment is described above in the method embodiment, and will not be repeated here.

[0085] In summary, by using the scheme of the device embodiment of the present disclosure, for the first entity to be processed, the corresponding entity in the knowledge base can be automatically determined through a series of processes, that is, entity linking is realized, and the entity linking model is not used, thereby saving the labor and time cost and ensuring the accuracy of the processing result.

[0086] The scheme of the present disclosure can be applied to the field of artificial intelligence, and particularly relates to the fields of knowledge graph, deep learning and natural language processing. Artificial intelligence is a discipline that studies enabling computers to simulate some thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.) of people. It has both hardware technology and software technology. The artificial intelligence hardware technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing, etc. The artificial intelligence software technology mainly includes computer vision technology, speech recognition technology, natural language processing technology, and machine learning / deep learning, big data processing technology, knowledge graph technology, etc.

[0087] In the technical scheme of the present disclosure, the entity of the character type is not for a specific user and cannot reflect the personal information of a specific user. In addition, the collection, storage, use, processing, transmission, provision and disclosure of user personal information comply with relevant laws and regulations and do not violate public order and good customs.

[0088] According to the embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0089] Figure 4A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0090] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 402 or a computer program loaded from storage unit 408 into random access memory (RAM) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0091] Multiple components in device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of monitors, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0092] The computing unit 401 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 401 performs various methods and processes described above, such as the methods described in the present disclosure. For example, in some embodiments, the methods described in the present disclosure can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded onto the RAM 403 and executed by the computing unit 401, one or more steps of the methods described in the present disclosure can be performed. Alternatively, in other embodiments, the computing unit 401 can be configured to perform the methods described in the present disclosure by any other appropriate means, such as by means of firmware.

[0093] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0094] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine or server, or entirely on a remote machine or server.

[0095] In the context of this disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0096] To provide for interaction with a user, the systems and techniques described here 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 a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, 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, speech, or tactile input.

[0097] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end 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), and the Internet.

[0098] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0099] It should be understood that the various forms of flow shown above can be used to reorder, add, or remove steps. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, so long as the desired results of the technology described in the present disclosure are achieved.

[0100] The specific implementation described above does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An entity linking method comprising: performing entity recognition on first text information corresponding to a first entity to be processed, including: performing a predetermined type of entity recognition on the first text information, the predetermined type being the same as a type of the first entity, the first text information including introduction information for the first entity; determining relevance between each recognized entity and the first entity using a pre-trained evaluation model, and sorting the recognized entities in descending order of the relevance, and filtering out entities other than the top M entities, M being a positive integer greater than 1; for any remaining entity after filtering, finding an entity corresponding to the entity from entities recorded in a knowledge base, and taking the found entity as a recall entity; obtaining second text information corresponding to each recall entity from the knowledge base, the second text information including introduction information for the recall entity; obtaining a relevance score between each second text information and the first text information, including: performing keyword extraction on the first text information, and determining relevance between each extracted keyword and the first text information using the evaluation model, assigning a corresponding weight to each extracted keyword according to the relevance, for any second text information, selecting keywords appearing in the second text information from the extracted keywords, and adding the corresponding weights of the selected keywords to obtain a sum, the sum being determined as the relevance score between the second text information and the first text information; and selecting an entity corresponding to the first entity from the recall entities according to the relevance score.

2. The method of claim 1, wherein the method further comprises: filtering out entities that are not of the predetermined type from the recall entities, and obtaining second text information corresponding to each remaining recall entity after filtering.

3. The method of claim 1, wherein, the obtaining of the relevance score between each second text information and the first text information further comprises: performing predetermined type keyword extraction on the first text information, and taking the extracted keywords as core keywords of the first text information; for any second text information, performing the predetermined type keyword extraction on the second text information, taking the extracted keywords as core keywords of the second text information, and determining the relevance score between the second text information and the first text information by comparing the core keywords of the second text information and the core keywords of the first text information.

4. The method of claim 1, wherein, the obtaining of the relevance score between each second text information and the first text information further comprises: performing keyword extraction on the first text information, for any second text information, selecting keywords appearing in the second text information from the extracted keywords, and determining a first relevance score between the second text information and the first text information according to the selection result; performing keyword extraction of a predetermined type on the first text information, taking the extracted keywords as core keywords of the first text information, for any second text information, performing keyword extraction of the predetermined type on the second text information respectively, taking the extracted keywords as core keywords of the second text information, and determining a second correlation score between the second text information and the first text information by comparing the core keywords of the second text information and the core keywords of the first text information; for any second text information, determining a final correlation score between the second text information and the first text information by combining the first correlation score and the second correlation score respectively.

5. The method of any one of claims 1-4, wherein, the selecting the entity corresponding to the first entity from the recall entities according to the correlation scores comprises: selecting a correlation score with the largest value from the correlation scores; if the correlation score with the largest value is greater than a predetermined threshold, taking the recall entity corresponding to the correlation score with the largest value as the entity corresponding to the first entity.

6. An entity linking apparatus comprising: an identification module, a recall module, an acquisition module, and a linking module; the identification module is configured to perform entity identification on first text information corresponding to a first entity to be processed, including: performing entity identification of a predetermined type on the first text information, the predetermined type being the same as the type of the first entity, the first text information including introduction information about the first entity; determining the relevance between each identified entity and the first entity by using a pre-trained evaluation model, and sorting the identified entities in descending order of the relevance, and filtering out entities other than the top M entities, M being a positive integer greater than 1; the recall module is configured to, for any entity remaining after filtering, find an entity corresponding to the entity from entities recorded in a knowledge base, and take the found entity as a recall entity; the acquisition module is configured to acquire second text information corresponding to each recall entity from the knowledge base, the second text information including introduction information about the recall entity; the linking module is configured to acquire a correlation score between each second text information and the first text information, including: performing keyword extraction on the first text information, and determining the relevance between each extracted keyword and the first text information by using an evaluation model, assigning a corresponding weight to each extracted keyword according to the relevance, for any second text information, selecting keywords appearing in the second text information from the extracted keywords, and adding the weights corresponding to the selected keywords to obtain a sum, and taking the sum as the correlation score between the second text information and the first text information; and selecting the entity corresponding to the first entity from the recall entities according to the correlation scores.

7. The apparatus of claim 6, wherein the recall module is further configured to filter out entities that are not of the predetermined type from the recall entities; The obtaining module respectively obtains second text information corresponding to each of the remaining recall entities after filtering.

8. The apparatus of claim 6, wherein, The linking module is further configured to perform predetermined type keyword extraction on the first text information, and take the extracted keywords as core keywords of the first text information; for any second text information, perform the predetermined type keyword extraction on the second text information respectively, and take the extracted keywords as core keywords of the second text information; and determine the relevance score between the second text information and the first text information by comparing the core keywords of the second text information and the core keywords of the first text information.

9. The apparatus of claim 6, wherein, The linking module is further configured to perform keyword extraction on the first text information; for any second text information, screen keywords appearing in the second text information from the extracted keywords respectively, and determine a first relevance score between the second text information and the first text information according to the screening result; perform predetermined type keyword extraction on the first text information, and take the extracted keywords as core keywords of the first text information; for any second text information, perform the predetermined type keyword extraction on the second text information respectively, and take the extracted keywords as core keywords of the second text information; and determine a second relevance score between the second text information and the first text information by comparing the core keywords of the second text information and the core keywords of the first text information; and for any second text information, determine a final relevance score between the second text information and the first text information by combining the first relevance score and the second relevance score respectively.

10. The apparatus of any one of claims 6-9, wherein, The linking module selects a relevance score with the largest value from the relevance scores, and takes the recall entity corresponding to the relevance score with the largest value as the entity corresponding to the first entity if the relevance score with the largest value is greater than a predetermined threshold.

11. An electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5. The computer instructions are used to enable the computer to perform the method of any one of claims 1-5.

13. A computer program product comprising computer programs / instructions, which, when executed by a processor, implement the method of any one of claims 1-5.

12. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, ​ ​

Images