Ontology-based method for constructing knowledge graph of potato industry chain

Abstract

The present invention belongs to the technical field of potato industry chain knowledge graph construction, and specifically relates to a method for constructing a potato industry chain knowledge graph based on ontology. It includes: ontology model construction of potato industry chain knowledge graph; acquisition of knowledge of each category in potato industry chain; construction of potato industry chain knowledge graph. The present invention analyzes the data characteristics of potato food industry chain knowledge graph and the efficient application characteristics of knowledge base, adopts artificial intelligence technology, natural language processing technology, knowledge engineering technology, automatic construction technology and organization and management technology for large-scale knowledge graph data, and studies the potato industry knowledge graph based on artificial intelligence, which has important theoretical significance for constructing a full-chain knowledge graph with industry characteristics.

Description

Technical Field

[0001] The present invention belongs to the technical field of potato industry chain knowledge graph construction, and specifically relates to an ontology-based potato industry chain knowledge graph construction method. Background Art

[0002] In recent years, with the rapid development of agricultural informatization, using knowledge graphs to form a large amount of scattered agricultural information into an interconnected whole, so that scholars and agricultural producers can obtain visual, concise and valuable information from a large amount of lengthy agricultural data, and guide farmers and technicians to make accurate decisions on agricultural production activities and agricultural industrialization, is a key issue that needs to be solved urgently. Therefore, it is very necessary to study the construction of agricultural knowledge bases. Early agricultural knowledge bases were mainly based on ontology construction. After years of research, a knowledge graph containing ontologies and many instances was formed. For example, in foreign countries, in 1980, the Food and Agriculture Organization and the European Commission jointly created a multilingual structured dictionary AGROVOC, covering related fields such as food, agriculture, forestry, and fisheries. In 2001, the agricultural ontology server AOS (Agricultural Ontology Service) was developed and applied. In China, the Agricultural Information Institute of the Chinese Academy of Agricultural Sciences took the lead in conducting research on agricultural knowledge based on ontology. The information classification, organization and intelligent reasoning provided by the project laid the foundation for the construction of agricultural intelligent search tools. At the same time, some researchers have proposed the idea of ​​constructing professional knowledge graphs for agricultural information of certain crops. Ontology construction models and successful cases have emerged in the fields of rice, flowers, and citrus, which have promoted the pace of research and application of knowledge graphs in the agricultural field and highlighted the fundamental position of knowledge graphs in the development of agricultural informatization and intelligence. In the future, the construction of agricultural information knowledge graphs will tend to be large-scale, combining ontology and examples.

[0003] However, in the field of agricultural production, there is currently a lack of open, large-scale, and full-industry chain knowledge graphs. In addition, there has been no research in China on the construction and representation methods of the knowledge graph for the entire industry chain in the potato field, let alone a knowledge graph representation model in this field. Potatoes are the fourth largest food production crop after rice, wheat, and corn. In order to avoid non-blind production in all links of the potato industry chain, including cultivation, planting, management, harvesting, and marketing, enhance knowledge and technical guidance, promote continuous and effective transition of all links, and improve production and income benefits, a large-scale, high-quality, and visualized potato industry chain knowledge graph is established. We strive to achieve new initiatives such as using knowledge databases to explore auxiliary potato variety selection, disease and pest prevention, and intelligent management of growth and processing processes. It has practical significance for effectively solving the problem that traditional knowledge bases are prone to failure in actual application scenarios, and has practical significance for making full use of information and knowledge to promote the refined, intelligent, and industrialized management of the potato industry, reduce its production costs, and improve its industrial efficiency. Summary of the invention

[0004] In view of the defects of the above-mentioned prior art, the present invention provides a method for constructing a potato industry chain knowledge graph based on ontology. The present invention analyzes the data characteristics of the potato food industry chain knowledge graph and the efficient application characteristics of the knowledge base, adopts artificial intelligence technology, natural language processing technology, knowledge engineering technology, automatic construction technology and organization and management technology for large-scale knowledge graph data, and studies the potato industry knowledge graph based on artificial intelligence, which has important theoretical significance for constructing a full-chain knowledge graph with industry characteristics.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] The ontology-based potato industry chain knowledge graph construction method includes:

[0007] S1. Construction of ontology model of potato industry chain knowledge graph;

[0008] S2, knowledge acquisition of various categories in the potato industry chain;

[0009] S3, construction of knowledge graph of potato industry chain;

[0010] S1 includes:

[0011] S101. Differentiate the whole potato industry chain into the following stages: the dormant germination period before production, the seedling growth period during production, and the harvest period after production;

[0012] S102. Analyze the potato industry categories, attributes, and relationships between categories and between categories and attributes, corresponding to each link of the potato industry chain, and construct an ontology model framework for the potato industry chain;

[0013] S2 includes:

[0014] S201, acquiring various types of knowledge in the potato industry chain based on the ontology model framework of the potato industry chain;

[0015] S202, cleaning, denoising and de-redundancy processing of various types of knowledge in the potato industry chain;

[0016] S203, extracting and processing various types of knowledge in the potato industry chain to form an instantiation model layer;

[0017] S3 includes:

[0018] Various types of knowledge in the potato industry chain are stored based on the graph database to form a knowledge graph of the potato industry chain.

[0019] The dormant germination period before production includes a knowledge framework of breeding, farming methods and regions;

[0020] Breeding includes breeding methods and variety names;

[0021] The farming method includes a description of the method and conditions;

[0022] The region includes soil and climate.

[0023] Knowledge framework for field management and pest and disease control during the seedling growth period;

[0024] Field management includes soil loosening, fertilization, irrigation, and weeding;

[0025] Among them, pest and disease control includes pest and disease characteristics and control methods.

[0026] The post-harvest period includes a knowledge framework for harvest storage, processing types and marketing;

[0027] Among them, harvest storage includes cold storage and cellaring;

[0028] The types of processing include food, starch, and feed;

[0029] Marketing includes direct sales, agency, import and export trade.

[0030] The various types of knowledge in the potato industry chain are: multi-source heterogeneous data of various structures obtained from various materials. The data are divided into structured, semi-structured and unstructured data. The method combining syntactic semantic analysis and weakly supervised learning is used to adaptively extract structured, semi-structured and unstructured data.

[0031] The instantiation pattern layer is formed in the following manner: semi-supervised knowledge fusion is used.

[0032] The beneficial effects of the present invention are as follows: 1) In view of the practical problems of discontinuous information and unrelated knowledge in the potato industry chain such as farming, planting, management and harvesting, an ontology layer model of the entire potato industry chain is defined based on professional field knowledge and data resources, and the data layer is instantiated into the ontology layer by combining natural language processing and deep learning technology, and a large-scale, high-quality, and visualized potato industry chain knowledge graph is constructed to realize the use of knowledge to assist in the intelligent management of potato variety breeding, disease and pest prevention, and growth and processing processes.

[0033] 2) An adaptive knowledge extraction method based on weakly supervised learning combined with syntactic and semantic analysis is proposed. This method focuses on the important features of potato food, uses lexical features, semantic role features, and context features to select entities, relationships, attributes, and events, and uses weakly supervised learning algorithms to iteratively obtain new templates for entities, relationships, and attributes. It can automatically extract multi-source heterogeneous data, break through the dependence of existing knowledge extraction methods on domains and structures, and improve the performance and automation of knowledge extraction.

[0034] 3) The present invention has practical significance for the refined, intelligent and industrialized management of the agricultural industry, reducing agricultural production costs and improving the efficiency of the agricultural industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is the basic framework diagram of the ontology model of the potato industry chain knowledge graph;

[0036] Figure 2 Construct a flow chart for the knowledge graph of the entire potato industry chain;

[0037] Figure 3 It is a relation extraction model that combines bidirectional LSTM with attention mechanism; DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific embodiments:

[0039] The preparation method and application of a WO3@NiCoP nano-adsorbent of the present invention are described in detail below through specific examples.

[0040] Example 1

[0041] (1) Construction of the ontology model of the potato industry chain knowledge graph

[0042] Based on the resource base and professional knowledge of the potato industry, combined with the characteristics of the potato's entire life cycle, the potato industry chain is defined as three links, namely the dormant germination period before production, the seedling growth period during production, and the harvest period after production. In these three periods, combined with the growth laws and main production factors of potatoes, the knowledge structure of the potato industry is deeply analyzed, and the core concept system of the potato industry chain is determined to be six categories: farming, planting, management, harvesting, processing, and sales. Further analysis of the potato industry categories, attributes, and the relationship between categories and between categories and attributes is conducted to construct a class chain structure diagram, such as Figure 1As shown in the figure, a logical relationship diagram of the knowledge system of the entire potato industry chain is formed, which is also the ontology model framework of the potato industry chain. The ontology layer is an open and integrated system that can effectively organize and manage a large amount of data dynamically and scalably, and deeply express the semantic relevance between knowledge. It provides a model guidance for the construction of the knowledge graph of the potato industry chain.

[0043] (2) Construction of the potato industry chain knowledge graph

[0044] The knowledge graph construction in this paper adopts a combination of bottom-up and top-down methods. First, based on a large amount of data, literature and other materials and professional knowledge, the ontology layer of the potato industry chain knowledge graph mentioned above is constructed. Then, multi-source heterogeneous data of various structures are obtained from websites such as the National Agricultural Science Data Center, China Potato Information Network, Baidu Encyclopedia and various literature materials. They are mainly divided into structured, semi-structured and unstructured data, and various types of data are cleaned, denoised, and de-redundant. Corresponding methods are used for various types of data to extract relevant knowledge of the potato industry chain, and then the model layer is instantiated. On the basis of the potato industry domain knowledge graph obtained by extracting examples, the ontology is generated and improved by completing the extraction tasks of domain concepts, hierarchical relationships between concepts, and non-hierarchical relationships, and the generated ontology is used to guide the extraction of instance knowledge, forming a technical closed loop of extracting ontology from extracted knowledge and then using ontology to guide knowledge extraction. The specific implementation process is as follows: Figure 2 As shown. Figure 2 In the process of building the knowledge graph of the potato industry chain, extracting entities and relationships from large amounts of unstructured data is a key technology that needs to be solved.

[0045] ① Knowledge extraction based on weak supervision

[0046] In view of the massive multi-source, heterogeneous, dynamic and diverse potato data, a method combining syntactic semantic analysis and weakly supervised learning is used to realize adaptive knowledge extraction. First, based on the above-constructed potato entity, attribute and relationship expression category system, the text data is subjected to syntactic semantic analysis with the objects in the potato industry chain ontology as the center, and candidate knowledge points related to potato objects are discovered. Then, lexical features, dependency syntactic features, semantic role features and context features are used to select entities and attributes layer by layer. The weakly supervised learning algorithm is used to iteratively train to obtain the patterns and categories of entities, attributes and relationships, so as to obtain "entity-relationship-entity" and "entity-attribute-attribute value" knowledge from massive data. This method uses a weakly supervised learning model based on the encoder-decoder attention mechanism to identify entities and extract relationships and attributes between entities. Specifically, Figure 3As shown. The relationship extraction of this model can automatically extract relationships by matching the domain knowledge base with a reliable text collection. The encoder of the relationship extraction model uses a bidirectional LSTM, which uses gated information selection to solve the gradient vanishing problem in long sequence learning and can obtain contextual semantic information of long texts. However, LSTM treats all input information equally and encodes it into information of a fixed length. The attention mechanism can treat input information differently in time and space, focusing on it, eliminating noise from irrelevant words and mislabeling, and improving the accuracy of entity relationship extraction.

[0047] ②Knowledge fusion based on semi-supervision

[0048] The chain information extracted by the above model is redundant, ambiguous and discrete. These fragmented information are treated as independent knowledge graphs, and the knowledge graphs are integrated through entity alignment and entity linking to solve the disambiguation and coreference resolution problems of various entity semantics such as homonyms, multiple names and abbreviations. For entity alignment, seed entities are first selected from multiple knowledge graphs to be fused and preprocessed. Then, dynamic indexing technology is used to index the attributes and relationships of the entities. After that, the pruning method is used to filter out entity pairs with low similarity and retain entity pairs with high similarity. Then, the attribute similarity and structural similarity methods are used to select matching entity pairs with high confidence with the seed entities to complete the entity alignment. For entity linking, the current knowledge graph is first used to generate a candidate entity set. According to the given entity link referent, the similarity between it and the candidate entity is calculated in combination with its contextual semantic features, and then sorted, and entity linking is realized according to the order.

[0049] (3) Knowledge graph construction

[0050] By storing the triplet knowledge extracted by the above method based on the graph database, the visual display and query of the knowledge graph can be realized. This paper adopts the Neo4j open source graph database system to store the triplet data of potato industry chain knowledge extracted by different methods into the Neo4j graph database to realize the visual display and knowledge management of the knowledge graph.

[0051] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

[0052] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for constructing a knowledge graph of the potato industry chain based on ontology, characterized in that: include: S1. Construction of ontology model of potato industry chain knowledge graph; S2, knowledge acquisition of various categories in the potato industry chain; S3, construction of knowledge graph of potato industry chain; S1 includes: S101. Differentiate the whole potato industry chain into the following stages: the dormant germination period before production, the seedling growth period during production, and the harvest period after production; S102. Analyze the potato industry categories, attributes, and relationships between categories and between categories and attributes, corresponding to each link of the potato industry chain, and construct an ontology model framework for the potato industry chain; S2 includes: S201, acquiring various types of knowledge in the potato industry chain based on the ontology model framework of the potato industry chain; S202, cleaning, denoising and de-redundancy processing of various types of knowledge in the potato industry chain; S203, extracting and processing various types of knowledge in the potato industry chain to form an instantiation model layer; S3 includes: Various types of knowledge in the potato industry chain are stored based on the graph database to form a knowledge graph of the potato industry chain.

2. The method for constructing a potato industry chain knowledge graph based on ontology according to claim 1 is characterized in that: The dormant germination period before production includes a knowledge framework of breeding, farming methods and regions; Breeding includes breeding methods and variety names; The farming method includes a description of the method and conditions; The region includes soil and climate.

3. The method for constructing a potato industry chain knowledge graph based on ontology according to claim 1 is characterized in that: Knowledge framework for field management and pest and disease control during the seedling growth period; Field management includes soil loosening, fertilization, irrigation, and weeding; Among them, pest and disease control includes pest and disease characteristics and control methods.

4. The method for constructing a potato industry chain knowledge graph based on ontology according to claim 1 is characterized in that: The post-harvest period includes a knowledge framework for harvest storage, processing types and marketing; Among them, harvest storage includes cold storage and cellaring; The types of processing include food, starch, and feed; Marketing includes direct sales, agency, import and export trade.

5. The method for constructing a potato industry chain knowledge graph based on ontology according to claim 1 is characterized in that: The various types of knowledge in the potato industry chain are: multi-source heterogeneous data of various structures obtained from various materials. The data are divided into structured, semi-structured and unstructured data. The method combining syntactic semantic analysis and weakly supervised learning is used to adaptively extract structured, semi-structured and unstructured data.

6. The method for constructing a potato industry chain knowledge graph based on ontology according to claim 1 is characterized in that: The instantiation pattern layer is formed in the following manner: semi-supervised knowledge fusion is used.

Images