Man-machine mixed knowledge graph construction method and system, electronic equipment and medium

Abstract

The invention provides a man-machine hybrid knowledge graph construction method and system, electronic equipment and a medium. The method comprises the following steps: constructing a specific domain data set based on a web crawler; designing a data set labeling algorithm based on a heuristic rule, and labeling the constructed specific field data set based on the web crawler by using the data set labeling algorithm based on the heuristic rule; constructing a specific domain data set based on expert experience; constructing a specific domain knowledge graph prototype through a deep neural network based on the specific domain data set; and constructing the man-machine hybrid knowledge graph based on the entity attributes and functions contained in the specific domain knowledge graph prototype. According to the technical scheme, the technical problems that time and labor are wasted and the coverage range is small due to the fact that the atlas is constructed manually in the specific field in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of human-machine hybrid intelligent technology, and in particular to a method, system, electronic device and medium for constructing a human-machine hybrid knowledge graph. Background Art

[0002] Knowledge is the information summarized by human beings in their objective practice of understanding the world. In the era of informatization and big data, as the sources of knowledge become more and more abundant, the amount of data is gradually increasing. Research on the means of representing knowledge and formal description is relatively scarce. Traditional data management and query methods are limited by the amount of data, and are also subject to certain constraints in terms of knowledge data in specific fields.

[0003] With the development of knowledge graphs, research on knowledge graphs in related fields is emerging, involving processes such as knowledge modeling, knowledge management and knowledge sharing. Knowledge modeling is a relatively accurate model representation of knowledge. Knowledge management is a process aimed at identifying, classifying and creating knowledge information, disseminating and sharing knowledge and transforming it into products, services and systems. Effective knowledge management is conducive to the use and continued generation of new knowledge, and ultimately achieves the purpose of knowledge sharing. Through research on knowledge representation, knowledge reasoning and knowledge graph construction, theoretical support can be provided for the development of knowledge graphs. The theoretical basis of knowledge graphs involves multiple disciplines, such as computer science, logic, semantics, etc. Knowledge graphs have been widely studied and applied in industries such as traditional Chinese medicine, ships, movies, music and even history. Ge Bin et al. proposed a set of construction processes from data collection, knowledge extraction to knowledge storage for graph construction, and determined knowledge representation models, storage frameworks and other contents for knowledge graph technology. Jiang Kai et al. constructed knowledge graphs and established databases for a large amount of knowledge in specific fields, and focused on detailed introduction of database queries. In view of the shortcomings of the communication field's complexity and lack of structure, Dai Jianwei and others built a knowledge base in the information and communication field, which can provide information data services in this field.

[0004] However, there are relatively few studies on the application of knowledge representation, knowledge reasoning, and graph construction technology in specific fields, and there is a lack of reference cases. The construction of knowledge graph technology in specific knowledge fields usually requires full use of expert experience and human intervention, and then manual construction. Although it meets the requirements in terms of accuracy, it consumes a lot of manpower and material resources and is difficult to cover a large amount of domain knowledge, and the description scope is limited. Summary of the invention

[0005] The present invention provides a method, system, electronic device and medium for constructing a human-machine hybrid knowledge graph, which can solve the technical problems in the prior art that manually constructing graphs in specific fields is time-consuming and labor-intensive and has a small coverage.

[0006] According to one aspect of the present invention, a method for constructing a human-machine hybrid knowledge graph is provided, the method comprising:

[0007] S1, construct a weapon and equipment attribute dataset based on web crawlers;

[0008] S2, design a dataset annotation algorithm based on heuristic rules, and use the dataset annotation algorithm based on heuristic rules to annotate the weapon equipment attribute dataset constructed in S1;

[0009] S3, construct a weapon equipment function dataset based on expert experience;

[0010] S4, based on the weapon and equipment attribute dataset and weapon and equipment function dataset, constructs a human-machine hybrid knowledge graph of weapon and equipment through a deep neural network.

[0011] Furthermore, S1 includes:

[0012] S11, by means of web crawlers, the Scrapy framework in Python is used to crawl various websites to obtain attribute text data of weapons and equipment;

[0013] S12, annotates the text data obtained in S11 by combining manual annotation with machine annotation, and generates a weapon equipment attribute data set through crawling, parsing and storage processes.

[0014] Furthermore, S2 includes:

[0015] S21, setting a labeled data set as a reference summary for ROUGE scoring, the labeled data set including sentences consisting of equipment types and relationship information;

[0016] S22, split the weapon equipment attribute dataset constructed in S1 into a sentence set, calculate the ROUGE score of each sentence in the sentence set and the sentence in the annotated dataset respectively, select sentences from the sentence set whose ROUGE score with the annotated dataset exceeds a certain threshold and add them to the corresponding sequence.

[0017] Furthermore, the ROUGE score of each sentence in the sentence set and the sentence in the annotated dataset is calculated by the following formula:

[0018]

[0019] In the above formula, S represents the sentence in the labeled dataset, gram n represents a word of length n in sentence S, Indicates the number of words of length n, It represents the number of times the word in sentence S appears in the input sentence input, and ROUGE-N(input) represents the ROUGE score of sentence S and input sentence input.

[0020] Furthermore, S4 includes:

[0021] S41, extracting data from the weapon equipment attribute data set and the weapon equipment function data set, and representing and storing them in a unified format after parsing and conversion;

[0022] S42, using the pre-trained named entity recognition model to identify entities in the data stored in S41, extracting the relationship between entity pairs in the data stored in S41 through the GRU model based on remote supervision, obtaining entities and corresponding entity relationships, and using the Neo4j graph database to store entity nodes and relationship edges;

[0023] S43, constructs a human-machine hybrid knowledge graph of weapons and equipment based on the entity nodes and relationship edges obtained in S42.

[0024] Furthermore, the weapon and equipment attribute data set includes classification information, attribute information and relationship information of various types of weapon and equipment.

[0025] Furthermore, the weapon and equipment function data set includes the application dimensions of various types of weapon and equipment.

[0026] According to another aspect of the present invention, a human-machine hybrid knowledge graph construction system is provided, the system comprising a data set construction module, a data set annotation module, a knowledge graph construction module and a human-machine hybrid knowledge graph construction module;

[0027] The dataset construction module is used to construct a weapon equipment attribute dataset based on web crawlers and a weapon equipment function dataset based on expert experience;

[0028] The data set annotation module is used to design a data set annotation algorithm based on heuristic rules, and use the data set annotation algorithm based on heuristic rules to annotate the weapon equipment attribute data set based on the web crawler;

[0029] The knowledge graph construction module is used to construct a human-machine hybrid knowledge graph of weapons and equipment through a deep neural network based on the weapons and equipment attribute dataset and the weapons and equipment function dataset.

[0030] According to another aspect of the present invention, an electronic device is provided, comprising a processor and a memory, wherein at least one instruction is stored in the memory, and the at least one instruction is loaded and executed by the processor to implement the method for constructing a human-machine hybrid knowledge graph as proposed in the present invention.

[0031] According to another aspect of the present invention, a computer-readable storage medium is provided, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement the human-computer hybrid knowledge graph construction method proposed in the present invention.

[0032] The technical solution of the present invention is applied to provide a method, system, electronic device and medium for constructing a human-machine hybrid knowledge graph. The method generates a data set by constructing a crawler, and develops a data set annotation algorithm to annotate the crawled data set, collects expert experience knowledge, and uses deep learning technology to construct a knowledge graph in a specific field. By distinguishing the static attributes and dynamic attributes of knowledge, a human-machine hybrid knowledge graph is formed; knowledge entities and their relationships are stored in a graph database to form a networked knowledge expression to support diverse knowledge retrieval and reasoning, and advanced knowledge graph technology is used to achieve efficient graph construction and visualization of professional knowledge in a specific field, so that staff can more intuitively understand the relationship, state and dynamic changes between various object entities, and provide important support for knowledge reasoning, collaborative decision-making and task planning. This method can effectively utilize existing object knowledge data information and provide more solid knowledge data resources for relevant personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The included drawings are used to provide a further understanding of the embodiments of the present invention, which constitute a part of the specification, are used to illustrate the embodiments of the present invention, and together with the text description, explain the principles of the present invention. Obviously, the drawings in the following description are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0034] Figure 1 A schematic diagram of the process of constructing a human-machine hybrid knowledge graph according to a specific embodiment of the present invention is shown;

[0035] Figure 2 A structural block diagram of a human-machine hybrid knowledge graph construction system provided according to a specific embodiment of the present invention is shown;

[0036] Figure 3 A structural block diagram of an electronic device provided according to a specific embodiment of the present invention is shown. DETAILED DESCRIPTION

[0037] It should be noted that, in the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments in 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.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] Unless otherwise specifically stated, the relative arrangement of the parts and steps described in these embodiments, numerical expressions and numerical values ​​do not limit the scope of the present invention. At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the accompanying drawings are not drawn according to the actual proportional relationship. The technology, method and equipment known to ordinary technicians in the relevant field may not be discussed in detail, but in appropriate cases, the technology, method and equipment should be regarded as a part of the authorization specification. In all examples shown and discussed here, any specific value should be interpreted as being merely exemplary, rather than as a limitation. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once a certain item is defined in an accompanying drawing, it does not need to be further discussed in subsequent drawings.

[0040] like Figure 1 As shown, according to a specific embodiment of the present invention, a method for constructing a human-machine hybrid knowledge graph is provided, and the method includes:

[0041] S1, construct a weapon and equipment attribute dataset based on web crawlers;

[0042] S2, design a dataset annotation algorithm based on heuristic rules, and use the dataset annotation algorithm based on heuristic rules to annotate the weapon equipment attribute dataset constructed in S1;

[0043] S3, construct a weapon equipment function dataset based on expert experience;

[0044] S4, based on the weapon and equipment attribute dataset and weapon and equipment function dataset, constructs a human-machine hybrid knowledge graph of weapon and equipment through a deep neural network.

[0045] By applying this configuration method, a method for constructing a human-machine hybrid knowledge graph is provided. The method generates a data set by constructing a crawler, and develops a data set annotation algorithm to annotate the crawled data set, collects expert experience knowledge, and uses deep learning technology to construct a knowledge graph in a specific field. By distinguishing the static attributes and dynamic attributes of knowledge, a human-machine hybrid knowledge graph is formed; knowledge entities and their relationships are stored in a graph database to form a networked knowledge expression to support diverse knowledge retrieval and reasoning, and advanced knowledge graph technology is used to achieve efficient graph construction and visualization of professional knowledge in a specific field, so that staff can more intuitively understand the relationship, state and dynamic changes between various object entities, and provide important support for knowledge reasoning, collaborative decision-making and task planning. This method can effectively utilize the existing object knowledge data information and provide more solid knowledge data resources for relevant personnel. Compared with the prior art, the technical solution of the present invention can solve the technical problems in the prior art that it is time-consuming and labor-intensive to manually construct graphs in specific fields and the coverage is small.

[0046] Furthermore, in the embodiment of the present invention, S1 includes:

[0047] S11, by means of web crawlers, the Scrapy framework in Python is used to crawl various websites to obtain attribute text data of weapons and equipment;

[0048] S12, annotates the text data obtained in S11 by combining manual annotation with machine annotation, and generates a weapon equipment attribute data set through crawling, parsing and storage processes.

[0049] Furthermore, in the embodiment of the present invention, S2 includes:

[0050] S21, setting a labeled data set as a reference summary for ROUGE scoring, the labeled data set including sentences consisting of equipment types and relationship information;

[0051] S22, split the weapon equipment attribute dataset constructed in S1 into a sentence set, calculate the ROUGE score of each sentence in the sentence set and the sentence in the annotated dataset respectively, select sentences from the sentence set whose ROUGE score with the annotated dataset exceeds a certain threshold and add them to the corresponding sequence.

[0052] As a specific embodiment of the present invention, the present invention calculates the ROUGE score of each sentence in the sentence set and the sentence in the annotated dataset respectively by the following formula:

[0053]

[0054] In the above formula, S represents the sentence in the labeled dataset, gram n represents a word of length n in sentence S, Indicates the number of words of length n, It represents the number of times the word in sentence S appears in the input sentence input, and ROUGE-N(input) represents the ROUGE score of sentence S and input sentence input, that is, the similarity measure between the two.

[0055] In practical applications, the present invention uses a text modeling and representation method of word vectors when performing heuristic rule-based data annotation, that is, when comparing sentence information, the sentence is broken down into individual words for similarity calculation to obtain a ROUGE score, that is, the value of n in the above ROUGE scoring formula is 1. Traditional models use word vector methods, but such models are affected by the quality of word segmentation. The use of word vectors does not require Chinese word segmentation operations on the text, thus avoiding the negative cascading effects of Chinese word segmentation in professional fields. This improves the accuracy of subsequent algorithms, and the use of word vectors can also reduce unregistered words.

[0056] Further, in an embodiment of the present invention, constructing a weapon equipment function data set based on expert experience includes:

[0057] Through manual form, the document data is screened and sorted to realize the extraction of manual experience;

[0058] Visit experts in related fields and use questionnaires, manual question-and-answer sessions, etc. to gain professional experience;

[0059] The collected experience is formatted and collated to obtain a weapon equipment function data set based on expert experience. In order to ensure quality, in the embodiment of the present invention, the formatted data is manually checked and the experience quality is verified as a specific field data set based on expert experience.

[0060] Based on the above embodiment, in the embodiment of the present invention, S4 includes:

[0061] S41, extracting data from the weapon equipment attribute data set and the weapon equipment function data set, and representing and storing them in a unified format after parsing and conversion;

[0062] S42, using the pre-trained named entity recognition model to identify entities in the data stored in S41, extracting the relationship between entity pairs in the data stored in S41 through the GRU model based on remote supervision, obtaining entities and corresponding entity relationships, and using the Neo4j graph database to store entity nodes and relationship edges;

[0063] Specifically, a pre-trained named entity recognition (NER) model is used to identify entities in the text, such as weapon types, technical terms, etc. A deep learning model is used for relationship extraction to extract the association between entities from the text. Named entity recognition (NER) is the process of extracting entities from text. We customized a labeling set containing various entities in the field of weapons and equipment, and used the Baidu LAC pre-trained model to perform fine-grained labeling on them to obtain entity type labels. On this basis, we used a BERT-based named entity recognition model to obtain a large amount of labeled data through remote supervision, and performed model training and parameter optimization. The optimized model achieved 92% F1 for named entity recognition on the test set, which is significantly better than the baseline model.

[0064] After identifying the entities, the relation extraction module extracts the semantic connection between entity pairs. We constructed a remote supervision dataset containing 10 different relation types, covering the main relations in the field of weapons and equipment, such as "equipment-performance parameters", "weapons-range", etc., to reduce the time and effort spent on manually constructing the dataset. Based on this dataset, we use BiLSTM based on the attention mechanism for relation classification. The Attention mechanism enables the model to focus on the keywords of the sentence and effectively extract relational features. At the same time, the attention mechanism is introduced at the sentence level. By calculating the clicks of the vector Q corresponding to each sentence and the vectors corresponding to other sentences, its attention weight is obtained, and the sentences with the highest information relevance are selected, which can effectively control the impact of redundant data on the experimental results. The current relation extraction model achieves an F1 score of 86%. Considering the mutual influence between entity recognition and relation extraction, we also tried a joint extraction method. This method uses a shared layer to learn entity and relation representations, and optimizes both simultaneously in a multi-task learning manner. The experimental results show that the joint model is slightly improved compared with independent training, with entity recognition F1 reaching 93% and relation extraction F1 reaching 88%.

[0065] In summary, through named entity recognition and relationship extraction, we can effectively extract the core elements of the knowledge graph from unstructured text. The extraction results provide structured information for building domain knowledge graphs and knowledge bases. The effect of entity and relationship extraction determines the upper limit of downstream tasks and is the basis for the success of knowledge graphs.

[0066] S43, constructs a human-machine hybrid knowledge graph of weapons and equipment based on the entity nodes and relationship edges obtained in S42.

[0067] On the one hand, it is aimed at algorithm testers, extracting relationships based on the remotely supervised GRU model and evaluating the final extraction results; on the other hand, it is aimed at users of the knowledge graph, showing the queried entities and the relationships between entities. The main functions of the knowledge graph prototype include: user management, data overview, entity query and relationship query.

[0068] Furthermore, in an embodiment of the present invention, the weapon equipment attribute data set includes classification information, attribute information and relationship information of various types of weapon equipment, which mainly describes the static attributes of the equipment itself. Classification information refers to the type of weapon equipment, such as firearms, aircraft, etc.; attribute information refers to parameters such as weight and range; relationship information such as "equipment-performance parameters", etc. The weapon equipment function data set includes the application dimensions of various types of weapon equipment. For example, "used for-task" mainly describes the application attributes of equipment. The two types of knowledge, one static and one dynamic, take into account some basic attribute capabilities of the object and the task activities involved, while also taking into account some specific application knowledge of the object during use. This multidimensional knowledge representation can more comprehensively reflect the characteristics of the entity and support complex reasoning and question-answering.

[0069] In the aforementioned equipment knowledge extraction process, based on the description of the "static attributes" of the object knowledge, the text preprocessing methods of word segmentation and part-of-speech tagging are adopted, and then the processed vocabulary is subjected to dependency syntactic analysis, and the single word segmentation results are formed into a "two-dimensional" semantic connection relationship, and then rule-based relationship extraction is performed on this to extract the object knowledge fact triples. The LTP graph construction tool and Python language are used for programming to realize entity extraction and relationship extraction.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] 1) The dataset is constructed based on web crawlers. Text data from well-known domestic and foreign websites are collected through web crawlers, and the corpus is annotated by combining manual annotation with machine annotation, which can solve the problem of large-scale acquisition and coverage of text data in specific fields;

[0072] 2) A dataset annotation algorithm based on heuristic rules can solve the problem that the existing Chinese automatic summarization dataset has no sentence-level labels, thus improving the annotation speed and accuracy;

[0073] 3) Complete the construction of the Chinese human expert experience database based on document review, questionnaires, etc., and use triples to represent expert experience, which can solve the formal representation of expert experience and form an expert experience database;

[0074] 4) For texts in specific fields, a text modeling and representation method using character distributed vectors is used. The method using word vectors in traditional models is affected by the quality of word segmentation, while the present invention uses word vectors and does not require Chinese word segmentation operations on texts. Therefore, the negative cascading effect of Chinese word segmentation in professional fields can be avoided, thereby improving the accuracy of subsequent algorithms. In addition, the use of word vectors can also reduce unregistered words. In addition, the number of Chinese characters is much smaller than the number of Chinese words, so the required memory can be reduced, thereby speeding up calculations.

[0075] 5) Using remote supervision to align a large amount of factual information in the existing knowledge base with the text of the training corpus to obtain a large-scale training dataset, which can avoid the time and effort spent on manually constructing the dataset;

[0076] 6) The GRU model is easier to converge due to its fewer parameters. Structurally, the GRU has only two gates (update and reset), which can directly pass the hidden state to the next unit. The present invention uses the GRU (Gate Recurrent Unit) neural network to extract text features, which can overcome the problem that traditional deep learning models cannot solve long-distance dependencies;

[0077] 7) Redundant data will have a negative impact on the experimental results. In order to avoid this, the present invention introduces an attention mechanism at the sentence level to accurately give the weight of each sentence in the set, thereby reducing the impact of such noise data on the experimental results;

[0078] 8) Adopt named entity recognition and relationship extraction models to identify entities in text and extract entity relationships, build a knowledge graph prototype, and use Neo4j graph database to store entity nodes and relationship edges, which can support the storage and use of large-scale node entities and relationships, and realize entity relationship query and other functions;

[0079] like Figure 2 As shown, according to another aspect of the present invention, a human-machine hybrid knowledge graph construction system is provided, the system comprising a data set construction module, a data set annotation module, a knowledge graph construction module and a human-machine hybrid knowledge graph construction module;

[0080] The dataset construction module is used to construct a weapon equipment attribute dataset based on web crawlers and a weapon equipment function dataset based on expert experience;

[0081] The data set annotation module is used to design a data set annotation algorithm based on heuristic rules, and use the data set annotation algorithm based on heuristic rules to annotate the weapon equipment attribute data set based on the web crawler;

[0082] The knowledge graph construction module is used to construct a human-machine hybrid knowledge graph of weapons and equipment through a deep neural network based on the weapons and equipment attribute dataset and the weapons and equipment function dataset.

[0083] Furthermore, in an embodiment of the present invention, the human-machine hybrid knowledge graph construction system also includes a data cleaning module for cleaning and preprocessing the data crawled by the web crawler, including removing duplicate data, processing missing values, normalizing data formats, etc., to ensure the accuracy and consistency of the data. In addition, the human-machine hybrid knowledge graph construction system also includes a data visualization module for converting the constructed complex knowledge graph and data set into intuitive and easy-to-understand charts, graphics and interactive interfaces by using data visualization tools and front-end technologies to achieve visual presentation.

[0084] According to another aspect of the present invention, an electronic device 300 is provided, the electronic device 300 includes a processor 301 and a memory 302, the memory 302 stores at least one instruction, and the at least one instruction is loaded and executed by the processor 301 to implement the human-machine hybrid knowledge graph construction method proposed in the present invention. In practical applications, the electronic device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 301 and one or more memories 302.

[0085] According to another aspect of the present invention, a computer-readable storage medium is provided, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement the method for constructing a human-computer hybrid knowledge graph as proposed in the present invention. As a specific embodiment of the present invention, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0086] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0087] In summary, the present invention provides a method, system, electronic device and medium for constructing a human-machine hybrid knowledge graph. The method generates a data set by constructing a crawler, and develops a data set annotation algorithm to annotate the crawled data set, collects expert experience knowledge, and uses deep learning technology to construct a knowledge graph in a specific field. By distinguishing the static attributes and dynamic attributes of knowledge, a human-machine hybrid knowledge graph is formed; knowledge entities and their relationships are stored in a graph database to form a networked knowledge expression to support diverse knowledge retrieval and reasoning, and advanced knowledge graph technology is used to achieve efficient graph construction and visualization of professional knowledge in a specific field, so that staff can more intuitively understand the relationship, state and dynamic changes between each object entity, and provide important support for knowledge reasoning, collaborative decision-making and task planning. Through this method, the existing object knowledge data information can be effectively utilized to provide relevant personnel with more solid knowledge data resources. Compared with the prior art, the technical solution of the present invention can solve the technical problems in the prior art that it is time-consuming and labor-intensive to manually construct graphs in specific fields and the coverage is small.

[0088] For ease of description, spatially relative terms such as "above", "above", "on the upper surface of", "above", etc. may be used here to describe the spatial positional relationship between a device or feature and other devices or features as shown in the figure. It should be understood that spatially relative terms are intended to include different orientations of the device in use or operation in addition to the orientation described in the figure. For example, if the device in the accompanying drawings is inverted, the device described as "above other devices or structures" or "above other devices or structures" will be positioned as "below other devices or structures" or "below other devices or structures". Thus, the exemplary term "above" can include both "above" and "below". The device can also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatially relative descriptions used here are interpreted accordingly.

[0089] In addition, it should be noted that the use of terms such as "first" and "second" to limit components is only for the convenience of distinguishing the corresponding components. If not otherwise stated, the above terms have no special meaning and therefore cannot be understood as limiting the scope of protection of the present invention.

[0090] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing a human-machine hybrid knowledge graph, characterized in that: The method comprises: S1, construct a weapon and equipment attribute dataset based on web crawlers; S2, designing a data set annotation algorithm based on heuristic rules, and annotating the weapon equipment attribute data set constructed in S1 by using the data set annotation algorithm based on heuristic rules; S3, construct a weapon equipment function dataset based on expert experience; S4, based on the weapon and equipment attribute dataset and weapon and equipment function dataset, constructs a human-machine hybrid knowledge graph of weapon and equipment through a deep neural network.

2. The method according to claim 1, characterized in that S1 includes: S11, by means of web crawlers, the Scrapy framework in Python is used to crawl various websites to obtain attribute text data of weapons and equipment; S12, annotates the text data obtained in S11 by combining manual annotation with machine annotation, and generates a weapon equipment attribute data set through crawling, parsing and storage processes.

3. The method according to claim 1, characterized in that S2 include: S21, setting a labeled data set as a reference summary for ROUGE scoring, wherein the labeled data set includes sentences consisting of equipment types and relationship information; S22, split the weapon equipment attribute data set constructed in S1 into a sentence set, calculate the ROUGE score of each sentence in the sentence set and the sentences in the annotated data set respectively, and select sentences from the sentence set whose ROUGE score with the annotated data set exceeds a certain threshold and add them to the corresponding sequence.

4. The method according to claim 3, characterized in that The ROUGE scores of each sentence in the sentence set and the sentences in the annotated dataset are calculated by the following formula: In the above formula, S represents the sentence in the labeled dataset, gram n represents a word of length n in sentence S, Indicates the number of words of length n, It represents the number of times the word in sentence S appears in the input sentence input, and ROUGE-N(input) represents the ROUGE score of sentence S and input sentence input.

5. The method according to claim 1, characterized in that S4 include: S41, extracting data from the weapon equipment attribute data set and the weapon equipment function data set, and representing and storing them in a unified format after parsing and conversion; S42, using the pre-trained named entity recognition model to identify entities in the data stored in S41, extracting the relationship between entity pairs in the data stored in S41 through the GRU model based on remote supervision, obtaining entities and corresponding entity relationships, and using the Neo4j graph database to store entity nodes and relationship edges; S43, constructs a human-machine hybrid knowledge graph of weapons and equipment based on the entity nodes and relationship edges obtained in S42.

6. The method according to any one of claims 1 to 5, characterized in that The weapon equipment attribute data set includes classification information, attribute information and relationship information of various types of weapon equipment.

7. The method according to any one of claims 1 to 5, characterized in that The weapon equipment function data set includes application dimensions of various types of weapon equipment.

8. A human-machine hybrid knowledge graph construction system, characterized in that: The system includes a data set construction module, a data set annotation module, a knowledge graph construction module, and a human-machine hybrid knowledge graph construction module; The data set construction module is used to construct a weapon equipment attribute data set based on a web crawler and a weapon equipment function data set based on expert experience; The data set annotation module is used to design a data set annotation algorithm based on heuristic rules, and use the data set annotation algorithm based on heuristic rules to annotate the weapon equipment attribute data set based on the web crawler; The knowledge graph construction module is used to construct a human-machine hybrid knowledge graph of weapons and equipment through a deep neural network based on a weapons and equipment attribute data set and a weapons and equipment function data set.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the human-machine hybrid knowledge graph construction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the human-machine hybrid knowledge graph construction method as described in any one of claims 1 to 7.

Images