Cross-language domain knowledge graph construction method and device
The method constructs a cross-language knowledge graph using seed vocabularies in multiple languages to expand and complement vocabularies, extract data, and perform preprocessing and fusion, addressing the limitations of existing methods by enhancing efficiency and reducing resource requirements.
Patent Information
- Application Number
- CN202011507796.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-12-18
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-12-18
AI Technical Summary
Current knowledge graph construction methods require raw data in a specific language and rely on expert knowledge and industry research, consuming significant manpower and material resources, and are unable to achieve cross-language knowledge graph construction.
A method and device for constructing a cross-language knowledge graph by obtaining seed vocabulary in at least two languages, expanding the vocabulary until a preset condition is met, performing cross-complementation to obtain relevant vocabularies, extracting original data, and constructing a knowledge graph through preprocessing, knowledge modeling, and fusion operations.
Enables cross-language knowledge graph construction without relying on expert knowledge, improving efficiency and reducing resource consumption, allowing for knowledge graph construction applicable to any field.
Smart Images

Figure CN112487213B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer technology, and in particular to a method and device for constructing a cross-language knowledge graph. Background Art
[0002] Knowledge graph is a database that stores knowledge. The concept of knowledge graph was formally proposed by Google in 2012. Its main purpose is to enhance search efficiency and improve user experience in the era of rapid development of the Internet and explosive growth of network data. With its excellent semantic processing technology and connectivity, knowledge graph has established a foundation for information intelligence applications. It is widely used in search, question and answer, intelligence analysis, etc., promoting the development of information technology from information services to knowledge services. In recent years, all walks of life have been studying the application of knowledge graphs in professional fields to better serve specific fields.
[0003] However, the current construction of knowledge graphs requires raw data in a certain language based on a specific field. It cannot achieve cross-language knowledge graph construction and requires the help of expert knowledge and industry research in that specific field, which consumes a lot of manpower and material resources. Summary of the invention
[0004] The present invention provides a method and device for constructing a cross-language knowledge graph, which is used to solve the defects that the construction of existing knowledge graphs needs to be based on original data of a certain language in a specific field, cannot realize cross-language knowledge graph construction, and requires the help of expert knowledge and industry research in the specific field to be realized, which will consume a lot of manpower and material resources.
[0005] The present invention provides a method for constructing a cross-language domain knowledge graph, comprising:
[0006] Obtain seed vocabulary in at least two languages corresponding to the target domain;
[0007] Using the seed words of the at least two languages, expanding the words until the number of words obtained by the expansion meets a preset condition, thereby obtaining the expanded words of the at least two languages;
[0008] Performing a cross-complementation operation on the extended vocabularies of the at least two languages to obtain relevant vocabularies in the target field;
[0009] Extracting original data corresponding to the relevant vocabulary from an existing database;
[0010] A knowledge graph is constructed for the original data to generate a knowledge graph of the target domain.
[0011] According to a cross-language domain knowledge graph construction method provided by the present invention, the cross-complementation operation is performed on the extended vocabulary of the at least two languages to obtain the relevant vocabulary of the target domain, including:
[0012] Obtaining intersection data of the extended vocabulary of the at least two languages, and using the intersection data as the relevant vocabulary of the target field; or,
[0013] Acquire union data of the extended vocabulary of the at least two languages, and use the union data as the relevant vocabulary of the target field; or,
[0014] The extended vocabulary of the at least two languages is prioritized, and the prioritized extended vocabulary is used as the relevant vocabulary of the target field.
[0015] According to a cross-language domain knowledge graph construction method provided by the present invention, the construction of the knowledge graph of the original data to generate the knowledge graph of the target domain includes:
[0016] Preprocessing the original data to obtain preprocessed target data;
[0017] Based on the preprocessed target data, a knowledge modeling operation is performed to obtain concept data, hierarchical relationships between different concepts, and concept attribute data;
[0018] Based on the preprocessed target data, a knowledge acquisition operation is performed to obtain instance data, a relationship between instances and concepts, and instance attribute data;
[0019] Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts and the instance attribute data, a knowledge fusion operation is performed to generate a knowledge graph of the target domain.
[0020] According to a cross-language domain knowledge graph construction method provided by the present invention, the knowledge modeling operation is performed based on the preprocessed target data to obtain concept data, hierarchical relationships between different concepts, and concept attribute data, including:
[0021] Based on the preprocessed target data, a concept acquisition operation is performed to obtain concept data;
[0022] Based on the preprocessed target data, a concept context generation operation is performed to obtain the hierarchical relationship between different concepts;
[0023] Based on the preprocessed target data, a concept attribute acquisition operation is performed to obtain concept attribute data.
[0024] According to a cross-language domain knowledge graph construction method provided by the present invention, the knowledge acquisition operation is performed based on the preprocessed target data to obtain instance data, the relationship between the instance and the concept, and the instance attribute data, including:
[0025] Based on the preprocessed target data, performing an instance extraction operation to obtain instance data;
[0026] Based on the preprocessed target data, performing instance classification operations to obtain the relationship between instances and concepts;
[0027] Based on the preprocessed target data, an instance attribute extraction operation is performed to obtain instance attribute data.
[0028] According to a cross-language domain knowledge graph construction method provided by the present invention, the knowledge fusion operation is performed according to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept and the instance attribute data to generate the knowledge graph of the target domain, including:
[0029] Based on the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, a concept fusion operation is performed to achieve concept layer data alignment;
[0030] According to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, an instance fusion operation is performed to achieve instance layer data alignment;
[0031] Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts, and the instance attribute data, a relationship fusion operation is performed to achieve alignment between concepts, between concepts and instances, and between instances, and to generate a knowledge graph for the target domain.
[0032] According to a cross-language domain knowledge graph construction method provided by the present invention, the preprocessing of the original data to obtain the preprocessed target data includes:
[0033] The original data is subjected to summary extraction, text extraction and / or information frame extraction to obtain pre-processed target data.
[0034] The present invention also provides a cross-language domain knowledge graph construction device, comprising:
[0035] A seed vocabulary acquisition unit, used to acquire seed vocabulary in at least two languages corresponding to the target domain;
[0036] A vocabulary expansion unit, configured to expand vocabulary using the seed vocabulary of the at least two languages until the number of expanded vocabulary meets a preset condition, thereby obtaining expanded vocabulary of the at least two languages;
[0037] A cross-supplementation unit, configured to perform a cross-supplementation operation on the extended vocabularies of the at least two languages to obtain relevant vocabularies in the target field;
[0038] An original data extraction unit, used for extracting original data corresponding to the relevant vocabulary from an existing database;
[0039] The knowledge graph construction unit is used to construct a knowledge graph for the original data to generate a knowledge graph of the target field.
[0040] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the cross-language domain knowledge graph construction method as described in any one of the above are implemented.
[0041] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for constructing a cross-language domain knowledge graph.
[0042] The cross-language domain knowledge graph construction method and device provided by the present invention utilize seed words in at least two languages of the target field to expand the vocabulary, cross-complement to obtain related words, obtain original data based on the related words, and construct the knowledge graph based on the original data, thereby realizing a cross-language domain knowledge graph construction method applicable to any field. The knowledge graph construction process does not need to rely on expert knowledge and industry research in the specific field, which can effectively improve the knowledge graph construction efficiency and save manpower and material resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0044] Figure 1 A schematic diagram of a process for constructing a cross-language domain knowledge graph provided by an embodiment of the present invention;
[0045] Figure 2A schematic diagram of the process of constructing a knowledge graph based on the original data to generate a knowledge graph of the target domain provided in an embodiment of the present invention;
[0046] Figure 3 A schematic diagram of the result of extracting the summary of "Yi Jianlian" provided by an embodiment of the present invention;
[0047] Figure 4 A schematic diagram of the result of extracting the "Yi Jianlian" abstract provided by an embodiment of the present invention;
[0048] Figure 5 A schematic diagram of the knowledge acquisition result of "Yi Jianlian" based on the summary extraction result provided by an embodiment of the present invention;
[0049] Figure 6 A schematic diagram of the "Yi Jianlian" knowledge acquisition result based on the summary extraction result provided by an embodiment of the present invention;
[0050] Figure 7 A schematic diagram of the "Yi Jianlian" page display provided by an embodiment of the present invention;
[0051] Figure 8 A schematic diagram showing the "Yi Jianlian" page provided in an embodiment of the present invention;
[0052] Fig. 9 A schematic diagram of the structure of a cross-language domain knowledge graph construction device provided by an embodiment of the present invention;
[0053] Fig.10 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0054] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0055] Combine the following Figure 1-Figure 10 The present invention describes the cross-language domain knowledge graph construction method and device.
[0056] First, the terms used in the present invention are explained.
[0057] Knowledge graph: A database that stores knowledge, which stores some triples, such as (Yao Ming, birthplace, Shanghai), etc. Each triple represents a fact. The knowledge graph can also be viewed as a graph. For example, in the above triple, Yao Ming and Shanghai are nodes, and birthplace is a directed and labeled edge from Yao Ming to Shanghai.
[0058] Concept: A type of entity in the knowledge graph, such as fruit, pear, etc.
[0059] Entity: A specific object in the knowledge graph, such as apple, hawthorn, etc.
[0060] Attributes: Characteristics of concepts or entities in the knowledge graph, such as the origin and color of an apple.
[0061] Relationship: The relationship between concepts, entities, and attributes in the knowledge graph and themselves or between any two of them. For example, the entity apple is one of the entities under the concept of fruit, and the color attribute of apple can be red, pink, golden yellow, etc.
[0062] Figure 1 A schematic diagram of a process for constructing a cross-language domain knowledge graph provided by an embodiment of the present invention includes:
[0063] Step 100: Obtain seed vocabulary in at least two languages corresponding to the target domain;
[0064] For a certain target domain, in order to construct a cross-language knowledge graph for the target domain, seed vocabulary in at least two languages of the target domain is first obtained.
[0065] Optionally, by receiving user input, in response to the user input, seed words of at least two languages included in the user input may be acquired.
[0066] Optionally, when the target domain is determined, seed vocabulary in at least two languages of the target domain may be directly acquired.
[0067] Step 101: Using the seed vocabulary of the at least two languages, vocabulary expansion is performed until the number of vocabulary obtained by expansion meets a preset condition, thereby obtaining the expanded vocabulary of the at least two languages;
[0068] The seed words in at least two languages of the target domain are used to expand the vocabulary through a pre-implemented vocabulary expansion function until the expanded vocabulary meets a preset condition, and finally the relevant words in at least two languages of the target domain are obtained.
[0069] Optionally, the expanded vocabulary meets the preset condition, which may be that the number of expanded vocabulary meets a preset number, or the ratio of the number of vocabulary of different languages in the expanded vocabulary meets a preset ratio, or other conditions that can be used to determine the end of vocabulary expansion.
[0070] Step 102: performing a cross-supplementation operation on the extended vocabulary of the at least two languages to obtain relevant vocabulary of the target field;
[0071] The extended vocabularies of the at least two languages are cross-complemented to improve and supplement the extended vocabularies of the at least two languages.
[0072] Optionally, performing a cross-complementation operation on the extended vocabulary of the at least two languages to obtain relevant vocabulary of the target field includes:
[0073] Obtaining intersection data of the extended vocabulary of the at least two languages, and using the intersection data as the relevant vocabulary of the target field; or,
[0074] Acquire union data of the extended vocabulary of the at least two languages, and use the union data as the relevant vocabulary of the target field; or,
[0075] The extended vocabulary of the at least two languages is prioritized, and the prioritized extended vocabulary is used as the relevant vocabulary of the target field.
[0076] Step 103: extracting original data corresponding to the relevant vocabulary from the existing database;
[0077] Corresponding raw data is extracted from an existing database according to relevant vocabulary in the target field.
[0078] An existing database is a database that already exists.
[0079] The original data is the source data of the related vocabulary. Optionally, the original data may be encyclopedia page information, web page original data, or other original data forms.
[0080] Step 104: construct a knowledge graph for the original data to generate a knowledge graph of the target domain.
[0081] The knowledge graph construction process includes preprocessing, knowledge modeling, knowledge acquisition, knowledge fusion and other steps.
[0082] Based on the original data, after preprocessing, knowledge modeling, knowledge acquisition, knowledge fusion and other steps, the knowledge graph of the target field is finally obtained.
[0083] In an embodiment of the present invention, by utilizing seed vocabularies in at least two languages of the target field to expand vocabulary and cross-complement related vocabulary, obtaining original data based on the related vocabulary, and constructing a knowledge graph based on the original data, a method for constructing a cross-language domain knowledge graph applicable to any field is implemented. The knowledge graph construction process does not need to rely on expert knowledge and industry research in the specific field, which can effectively improve the construction efficiency of the knowledge graph and save manpower and material resources.
[0084] Based on the above embodiment, optionally, Figure 2 As shown, the construction of the knowledge graph based on the original data to generate the knowledge graph of the target field includes:
[0085] Step 200: preprocess the original data to obtain preprocessed target data;
[0086] Preprocessing refers to normalizing data before the main processing.
[0087] Optionally, in one embodiment, preprocessing the original data to obtain preprocessed target data includes:
[0088] The original data is subjected to summary extraction, text extraction and information frame extraction to obtain pre-processed target data.
[0089] Step 201: Based on the pre-processed target data, perform a knowledge modeling operation to obtain concept data, hierarchical relationships between different concepts, and concept attribute data;
[0090] Knowledge modeling includes the processes of concept acquisition, concept hypernymy generation, and concept attribute acquisition. Concept acquisition is to extract concept data from preprocessed data, concept hypernymy generation is to obtain the hypernymy relationship between different concepts from the original data through certain rules, and concept attribute extraction is the process of extracting concept attributes.
[0091] Optionally, in one embodiment, the performing of knowledge modeling operations based on the preprocessed target data to obtain concept data, hierarchical relationships between different concepts, and concept attribute data includes:
[0092] Based on the preprocessed target data, a concept acquisition operation is performed to obtain concept data;
[0093] Based on the preprocessed target data, a concept context generation operation is performed to obtain the hierarchical relationship between different concepts;
[0094] Based on the preprocessed target data, a concept attribute acquisition operation is performed to obtain concept attribute data.
[0095] Step 202: Based on the preprocessed target data, perform a knowledge acquisition operation to obtain instance data, the relationship between the instance and the concept, and instance attribute data;
[0096] Knowledge acquisition mainly includes instance extraction, instance classification, instance attribute extraction, etc. Instance extraction is to extract instance data from preprocessed data, instance classification is to extract the relationship between instance and concept from preprocessed data, and instance attribute extraction is to extract instance attribute data from preprocessed data.
[0097] Optionally, in one embodiment, performing a knowledge acquisition operation based on the preprocessed target data to obtain instance data, a relationship between an instance and a concept, and instance attribute data includes:
[0098] Based on the preprocessed target data, performing an instance extraction operation to obtain instance data;
[0099] Based on the preprocessed target data, performing instance classification operations to obtain the relationship between instances and concepts;
[0100] Based on the preprocessed target data, an instance attribute extraction operation is performed to obtain instance attribute data.
[0101] Step 203: Perform a knowledge fusion operation based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts, and the instance attribute data to generate a knowledge graph for the target domain.
[0102] Knowledge fusion mainly includes concept fusion, instance fusion and relationship fusion. Concept fusion mainly refers to the fusion of concept-level data, instance fusion mainly refers to the fusion of instance-level data, and relationship fusion refers to the fusion of the relationship between concepts, the relationship between concepts and instances, and the relationship between instances.
[0103] Optionally, in one embodiment, performing a knowledge fusion operation to generate a knowledge graph of the target domain according to the concept data, the hierarchical relationships between different concepts and the concept attribute data, and the instance data, the relationship between the instance and the concept, and the instance attribute data includes:
[0104] Based on the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, a concept fusion operation is performed to achieve concept layer data alignment;
[0105] According to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, an instance fusion operation is performed to achieve instance layer data alignment;
[0106] Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts, and the instance attribute data, a relationship fusion operation is performed to achieve alignment between concepts, between concepts and instances, and between instances, and to generate a knowledge graph for the target domain.
[0107] In an embodiment of the present invention, raw data is extracted through vocabulary expansion and cross-complementation of a small number of seed words in at least two languages in a certain field, and a knowledge graph is constructed based on the raw data. The upper conceptual layer data can be constructed without expert knowledge. A small number of cross-domain seed words in a certain field can quickly realize the construction of a cross-language knowledge graph in the field, which can effectively improve the construction efficiency of the cross-language knowledge graph.
[0108] A specific example is given below to further illustrate the cross-language domain knowledge graph construction method provided by the present invention.
[0109] Construct a cross-language basketball-related knowledge graph in Chinese and English. First, input 20 Chinese seed words such as "National Basketball League, Yao Ming, Kobe, Basketball" and 20 English seed words such as "National Basketball Association, YaoMing, Kobe, Basketball"; then, through the vocabulary expansion function, the expansion function of the Chinese basketball field vocabulary in Chinese and English was realized respectively, among which 450 related words were obtained in Chinese, such as "Yi Jianlian, Sun Yue, Chinese Men's Basketball Professional League, James, Lakers, Champion, Wade, Heat" and other words, and 403 related words were obtained in English, such as "Yi Jianlian, Sun Yue, National Basketball Association, James, Lakers" and other words; then, a cross-supplement operation was performed, and the corresponding Chinese and English words were combined to obtain 455 related words in Chinese and English; according to the related words, the corresponding original data was extracted from the database, such as the encyclopedia page information corresponding to the related words, and then the knowledge graph was constructed.
[0110] The knowledge graph construction process includes: first, preprocessing the original data, including operations such as abstract extraction, text extraction and / or information box extraction; the result of the abstract extraction of "Yi Jianlian" is as follows: Figure 3As shown, the results of the "YiJianlian" summary are Figure 4 shown.
[0111] On the basis of preprocessing, the knowledge modeling process is carried out. Knowledge modeling includes the processes of concept acquisition, concept upper and lower generation, and concept attribute acquisition. Among them, concept acquisition is to extract concept data from the preprocessed data, concept upper and lower generation is to obtain the upper and lower relationship between different concepts from the original data through certain rules, and concept attribute extraction is the process of extracting concept attributes. Among them, the concepts in the Chinese part include "athlete", "rules", "league", etc., among which "Yi Jianlian" is the lower instance of the concept of "athlete", and the concept of "athlete" has attributes such as "position on the field", "birth time", and "nationality". The concepts in the English part include "Players", "Rules", "League", etc., among which "YiJianlian" is the lower instance of the concept of "Players", and the concept of "Players" has attributes such as "Position", "Born", and "Nationality".
[0112] Then, knowledge acquisition is performed. Knowledge acquisition mainly includes instance extraction, instance classification, instance attribute extraction, etc. Instance extraction is to extract instance data from preprocessed data, instance classification is to extract the relationship between instance and concept from preprocessed data, and instance attribute extraction is to extract instance attribute data from preprocessed data. Figure 5 A schematic diagram of the knowledge acquisition result of "Yi Jianlian" based on the summary extraction result provided in the embodiment of the present application, Figure 6 A schematic diagram of the "Yi Jianlian" knowledge acquisition results based on the summary extraction results provided in an embodiment of the present application.
[0113] Through the above steps of vocabulary expansion, preprocessing, knowledge modeling, knowledge acquisition, etc., we successfully constructed basketball-related knowledge graph data, including 5 concepts, 455 instances, and 220 attributes in Chinese. There are 5 concepts, 455 instances, and 210 attributes in English. Finally, the Chinese and English data are fused, and the concept layer data, example layer data, and attribute alignment are carried out respectively. In this example, the concept layer alignment results are: "Players", "Rules", "League", etc., the instance layer alignment results are: "Yi Jianlian", "Sun Yue", "James", "Lakers", "Champion", and the attribute alignment results are: "Position", "Birth time", "Nationality", etc.
[0114] Finally, we obtain the cross-language knowledge graph of the target domain. All the obtained triples are stored in the database virtuoso, and then displayed on the page. The interface of "Yi Jianlian" and "Yi Jianlian" is as follows Figure 7 and Figure 8 As shown, the two interfaces can be switched via CN (Chinese) and EN (English).
[0115] The cross-language domain knowledge graph construction device provided by the present invention is described below. The cross-language domain knowledge graph construction device described below and the cross-language domain knowledge graph construction method described above can be referenced to each other.
[0116] Fig. 9 The schematic diagram of the structure of the cross-language domain knowledge graph construction device provided by the embodiment of the present invention includes: a seed vocabulary acquisition unit 910, a vocabulary expansion unit 920, a cross-supplementation unit 930, a raw data extraction unit 940 and a knowledge graph construction unit 950, wherein:
[0117] A seed vocabulary acquisition unit 910 is used to acquire seed vocabulary in at least two languages corresponding to the target domain;
[0118] A vocabulary expansion unit 920 is used to expand vocabulary using the seed vocabulary of the at least two languages until the number of expanded vocabulary meets a preset condition, thereby obtaining expanded vocabulary of the at least two languages;
[0119] A cross-supplementation unit 930, configured to perform a cross-supplementation operation on the extended vocabularies of the at least two languages to obtain relevant vocabularies in the target field;
[0120] The original data extraction unit 940 is used to extract the original data corresponding to the relevant words from the existing database;
[0121] The knowledge graph construction unit 950 is used to construct a knowledge graph for the original data to generate a knowledge graph of the target domain.
[0122] Optionally, the cross-complementation unit 930 is used for:
[0123] Obtaining intersection data of the extended vocabulary of the at least two languages, and using the intersection data as the relevant vocabulary of the target field; or,
[0124] Acquire union data of the extended vocabulary of the at least two languages, and use the union data as the relevant vocabulary of the target field; or,
[0125] The extended vocabulary of the at least two languages is prioritized, and the prioritized extended vocabulary is used as the relevant vocabulary of the target field.
[0126] Optionally, the knowledge graph construction unit 950 includes:
[0127] A preprocessing submodule, used to preprocess the raw data to obtain preprocessed target data;
[0128] A knowledge modeling submodule is used to perform knowledge modeling operations based on the preprocessed target data to obtain concept data, hierarchical relationships between different concepts, and concept attribute data;
[0129] A knowledge acquisition submodule, used to perform a knowledge acquisition operation based on the preprocessed target data to obtain instance data, the relationship between the instance and the concept, and instance attribute data;
[0130] The knowledge fusion submodule is used to perform knowledge fusion operations based on the concept data, the hierarchical relationships between different concepts and concept attribute data, as well as the instance data, the relationship between instances and concepts and instance attribute data to generate the knowledge graph of the target domain.
[0131] Optionally, the knowledge modeling submodule is used to:
[0132] Based on the preprocessed target data, a concept acquisition operation is performed to obtain concept data;
[0133] Based on the preprocessed target data, a concept context generation operation is performed to obtain the hierarchical relationship between different concepts;
[0134] Based on the preprocessed target data, a concept attribute acquisition operation is performed to obtain concept attribute data.
[0135] Optionally, the knowledge acquisition submodule is used to:
[0136] Based on the preprocessed target data, performing an instance extraction operation to obtain instance data;
[0137] Based on the preprocessed target data, performing instance classification operations to obtain the relationship between instances and concepts;
[0138] Based on the preprocessed target data, an instance attribute extraction operation is performed to obtain instance attribute data.
[0139] Optionally, the knowledge fusion submodule is used to:
[0140] Based on the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, a concept fusion operation is performed to achieve concept layer data alignment;
[0141] According to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, an instance fusion operation is performed to achieve instance layer data alignment;
[0142] Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts, and the instance attribute data, a relationship fusion operation is performed to achieve alignment between concepts, between concepts and instances, and between instances, and to generate a knowledge graph for the target domain.
[0143] Optionally, the preprocessing submodule is used to:
[0144] The original data is subjected to summary extraction, text extraction and information frame extraction to obtain pre-processed target data.
[0145] The cross-language domain knowledge graph construction device provided by the present invention can realize Figures 1 to 8 The various processes implemented by the method embodiment and achieving the same technical effect are not described here to avoid repetition.
[0146] Fig.10 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.10As shown, the electronic device may include: a processor 1010, a communication interface 1020, a memory 1030 and a communication bus 1040, wherein the processor 1010, the communication interface 1020 and the memory 1030 communicate with each other through the communication bus 1040. The processor 1010 may call the logic instructions in the memory 1030 to execute a cross-language domain knowledge graph construction method, the method comprising: obtaining seed words in at least two languages corresponding to the target domain; using the seed words in the at least two languages to perform vocabulary expansion until the number of words obtained by the expansion meets the preset conditions, and obtaining the expanded words in the at least two languages; performing cross-supplementation operations on the expanded words in the at least two languages to obtain the relevant words in the target domain; extracting the original data corresponding to the relevant words from the existing database; constructing a knowledge graph on the original data to generate the knowledge graph of the target domain.
[0147] In addition, the logic instructions in the above-mentioned memory 1030 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.
[0148] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the cross-language domain knowledge graph construction method provided by the above methods, and the method includes: obtaining seed vocabulary in at least two languages corresponding to the target domain; using the seed vocabulary in the at least two languages to expand the vocabulary until the number of expanded vocabulary meets a preset condition, and obtaining the expanded vocabulary in the at least two languages; performing cross-complement operations on the expanded vocabulary in the at least two languages to obtain relevant vocabulary in the target domain; extracting the original data corresponding to the relevant vocabulary from an existing database; constructing a knowledge graph for the original data to generate a knowledge graph for the target domain.
[0149] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the above-mentioned cross-language domain knowledge graph construction method, the method comprising: obtaining seed vocabulary in at least two languages corresponding to the target domain; using the seed vocabulary in the at least two languages to expand vocabulary until the number of expanded vocabulary meets a preset condition, thereby obtaining the expanded vocabulary in the at least two languages; performing cross-complement operations on the expanded vocabulary in the at least two languages to obtain relevant vocabulary in the target domain; extracting original data corresponding to the relevant vocabulary from an existing database; constructing a knowledge graph for the original data to generate a knowledge graph for the target domain.
[0150] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0151] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for constructing a cross-language domain knowledge graph, characterized in that: include: Obtain seed vocabulary in at least two languages corresponding to the target domain; Using the seed words of the at least two languages, expanding the words until the number of words obtained by the expansion meets a preset condition, thereby obtaining the expanded words of the at least two languages; Performing a cross-complementation operation on the extended vocabularies of the at least two languages to obtain relevant vocabularies in the target field; Extracting original data corresponding to the relevant vocabulary from an existing database; Constructing a knowledge graph for the original data to generate a knowledge graph for the target domain; The performing a cross-complementation operation on the extended vocabulary of the at least two languages to obtain relevant vocabulary of the target field includes: Obtaining intersection data of the extended vocabulary of the at least two languages, and using the intersection data as the relevant vocabulary of the target field; or, Acquire union data of the extended vocabulary of the at least two languages, and use the union data as the relevant vocabulary of the target field; or, Prioritizing the extended vocabulary in the at least two languages, and using the prioritized extended vocabulary as relevant vocabulary in the target field; The constructing of the knowledge graph of the original data to generate the knowledge graph of the target domain includes: Preprocessing the original data to obtain preprocessed target data; Based on the preprocessed target data, a knowledge modeling operation is performed to obtain concept data, hierarchical relationships between different concepts, and concept attribute data; Based on the preprocessed target data, a knowledge acquisition operation is performed to obtain instance data, a relationship between instances and concepts, and instance attribute data; Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts and the instance attribute data, a knowledge fusion operation is performed to generate a knowledge graph of the target domain.
2. The cross-language domain knowledge graph construction method according to claim 1, characterized in that: The performing of knowledge modeling operations based on the preprocessed target data to obtain concept data, hierarchical relationships between different concepts, and concept attribute data includes: Based on the preprocessed target data, a concept acquisition operation is performed to obtain concept data; Based on the preprocessed target data, a concept context generation operation is performed to obtain the hierarchical relationship between different concepts; Based on the preprocessed target data, a concept attribute acquisition operation is performed to obtain concept attribute data.
3. The cross-language domain knowledge graph construction method according to claim 1, characterized in that: The performing of knowledge acquisition operation based on the preprocessed target data to obtain instance data, the relationship between the instance and the concept, and instance attribute data includes: Based on the preprocessed target data, performing an instance extraction operation to obtain instance data; Based on the preprocessed target data, performing instance classification operations to obtain the relationship between instances and concepts; Based on the preprocessed target data, an instance attribute extraction operation is performed to obtain instance attribute data.
4. The cross-language domain knowledge graph construction method according to claim 1, characterized in that: The performing of knowledge fusion operation according to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept and the instance attribute data to generate the knowledge graph of the target domain includes: Based on the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, a concept fusion operation is performed to achieve concept layer data alignment; According to the concept data, the hierarchical relationship between different concepts and the concept attribute data, as well as the instance data, the relationship between the instance and the concept, and the instance attribute data, an instance fusion operation is performed to achieve instance layer data alignment; Based on the concept data, the hierarchical relationships between different concepts and the concept attribute data, as well as the instance data, the relationship between instances and concepts, and the instance attribute data, a relationship fusion operation is performed to achieve alignment between concepts, between concepts and instances, and between instances, and to generate a knowledge graph for the target domain.
5. The method for constructing a cross-language domain knowledge graph according to claim 1, characterized in that: The preprocessing of the original data to obtain the preprocessed target data includes: The original data is subjected to summary extraction, text extraction and / or information frame extraction to obtain pre-processed target data.
6. A cross-language domain knowledge graph construction device, characterized in that: include: A seed vocabulary acquisition unit, used to acquire seed vocabulary in at least two languages corresponding to the target domain; A vocabulary expansion unit, configured to expand vocabulary using the seed vocabulary of the at least two languages until the number of expanded vocabulary meets a preset condition, thereby obtaining expanded vocabulary of the at least two languages; A cross-supplementation unit, configured to perform a cross-supplementation operation on the extended vocabularies of the at least two languages to obtain relevant vocabularies in the target field; An original data extraction unit, used for extracting original data corresponding to the relevant vocabulary from an existing database; A knowledge graph construction unit, used to construct a knowledge graph for the original data to generate a knowledge graph of the target domain; The cross-complementary unit is used to: Acquire intersection data of the extended vocabulary of the at least two languages, and use the intersection data as the relevant vocabulary of the target field; or, Acquire union data of the extended vocabulary of the at least two languages, and use the union data as the relevant vocabulary of the target field; or, Prioritizing the extended vocabulary in the at least two languages, and using the prioritized extended vocabulary as relevant vocabulary in the target field; The knowledge graph construction unit includes: A preprocessing submodule, used to preprocess the raw data to obtain preprocessed target data; A knowledge modeling submodule is used to perform knowledge modeling operations based on the preprocessed target data to obtain concept data, hierarchical relationships between different concepts, and concept attribute data; A knowledge acquisition submodule, used to perform a knowledge acquisition operation based on the preprocessed target data to obtain instance data, the relationship between the instance and the concept, and instance attribute data; The knowledge fusion submodule is used to perform knowledge fusion operations based on the concept data, the hierarchical relationships between different concepts and concept attribute data, as well as the instance data, the relationship between instances and concepts and instance attribute data to generate the knowledge graph of the target domain.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for constructing a cross-language domain knowledge graph as described in any one of claims 1 to 5 are implemented.
8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing a cross-language domain knowledge graph as described in any one of claims 1 to 5 are implemented.
Citation Information
Patent Citations
Knowledge graph generating method and device
CN106156365A
Knowledge graph construction method and device, computer equipment and storage medium
CN110750698A
Cross-language medical knowledge graph construction method and device and electronic equipment
CN110990579A
Semantic knowledge base
US20190311003A1
