An ontology modeling method and modeling system for flower diseases and insect pests based on a knowledge graph

Abstract

The application discloses a kind of based on knowledge graph's ontology modeling method of flower disease and insect pest.The ontology modeling method constructs the ontology model capable of meeting the basic requirements of flower disease and insect pest control by studying field-related literature, compared with other existing ontology, additionally considers the influence of environment on flower disease and insect pest control, environmental factors for the prevention and treatment of flower disease and insect pest not only focus on treatment, more focus on prevention, timely prevention and treatment of disease and insect pest can further reduce the damage of flower.Using RDF graph to store ontology structure, according to the self-defined RDF2PG mapping method, the extracted triples are not stored in Neo4j database according to the structure of ontology model without other storage methods, which standardizes the managed knowledge, improves the storage efficiency and automatic construction of atlas ability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of flower disease and pest control, and particularly relates to an ontology modeling method and a modeling system for flower diseases and pests based on a knowledge graph. BACKGROUND

[0002] Flower diseases and pests are important reasons for low flower production efficiency. If the diseases and pests occurring in the flower planting process cannot be handled in time, the planting income of flowers will be seriously affected. With the development of information technology, various flower disease and pest control knowledge is scattered in the network. The traditional relational database knowledge management method cannot effectively represent and store these knowledge, and there are problems such as inability to fuse heterogeneous data, inability to efficiently express the relationship between data, and inability to refine knowledge. At present, the research in the field of flower diseases and pests mainly focuses on the prevention and control strategies, prevention and control knowledge, and occurrence principles of a certain type of flower diseases and pests, and lacks the sorting and integration of entity relationships in flower disease and pest knowledge, and misses the correlation between flower disease and pest entities, resulting in scattered and redundant flower disease and pest knowledge, and lack of disease and pest knowledge management and modeling tools and good methods.

[0003] In the field of knowledge graph research on diseases and pests, some experts and scholars have made some achievements. Some scholars focus on the research of domain ontology, reference agricultural thesaurus and disease and pest related literature, construct a disease and pest domain ontology model, and solve practical problems through the ontology model. Some other scholars focus on the research of domain data, analyze the domain data, extract entities and relationships in the data through knowledge extraction method, and realize the refinement of knowledge. Some other scholars use bibliometric method, construct knowledge graph through keyword clustering, and visually display the research in the field of diseases and pests.

[0004] These studies ignore the environment, which is crucial for disease and pest control, in the sorting of disease and pest knowledge, lack intelligence and systematization in management, and have great improvement in the generalization and accuracy of knowledge graph construction method and unstructured data content extraction method. The research in the field of flower diseases and pests mainly focuses on the prevention and control strategies, prevention and control knowledge, and occurrence principles of a certain type of flower diseases and pests, and lacks the sorting and integration of entity relationships in flower disease and pest knowledge, and misses the correlation between flower disease and pest entities, resulting in scattered and redundant flower disease and pest knowledge. SUMMARY

[0005] The present application aims to at least solve one of the above technical problems or at least provide a useful commercial choice. To this end, one object of the present application is to propose an ontology modeling method for flower pests and diseases based on a knowledge graph, which extracts flower pests and diseases prevention elements including the environment, builds a flower pests and diseases ontology model by reusing existing pests and diseases knowledge systems, and stores it using a resource description framework graph. After analyzing the flower pests and diseases literature corpus, the head-tail entity separation "01" annotation method is used to solve the annotation problem of nested head-tail entities, the ALBERT pre-training model is used for semantic feature extraction, and the CasPOSRel model combining the part-of-speech feature vector and the hierarchical annotation model (CasRel) is proposed to jointly extract triples in a large amount of flower pests and diseases text. At the same time, according to the ontology model constructed, the self-defined RDF2PG mapping method is used to store the extracted triples in the Neo4j graph database according to the ontology structure in the resource description framework graph, complete the storage and management of flower pests and diseases knowledge, and apply knowledge discovery to find the most susceptible environment for various types of flowers, thereby playing a preventive role for pests and diseases. The modeling method proposed in this paper can support intelligent diagnosis, decision-making, and question answering for flower pests and diseases, and improve the efficiency and production level of flower pests and diseases prevention. Another object of the present application is to propose an ontology modeling system for flower pests and diseases based on a knowledge graph.

[0006] The ontology modeling method for flower pests and diseases based on a knowledge graph according to the present application comprises the following steps:

[0007] extracting a plurality of attribute elements in the field of flower pests and diseases from the text;

[0008] constructing an ontology model in the field of flower pests and diseases, wherein the ontology model comprises a triple unit;

[0009] annotating a head entity array and a tail entity array in the triple unit, respectively;

[0010] constructing a joint extraction framework model based on the head entity array, the tail entity array, and the relationship between the head entity array and the tail entity array;

[0011] establishing a knowledge extraction framework based on a knowledge graph using a pre-trained language representation model;

[0012] convert the resource description framework in the triple unit into an attribute graph, and store the attribute graph in a Neo4j graph database.

[0013] Compared with other existing ontologies, the ontology modeling method for flower diseases and insect pests based on a knowledge graph additionally considers the influence of the environment on the prevention and treatment of flower diseases and insect pests, and the environmental factors are not only focused on treatment, but also focused on prevention, and timely prevention of diseases and insect pests can further reduce the damage to flowers.

[0014] In addition, the ontology modeling method for flower diseases and insect pests based on a knowledge graph according to the present applicationapplicationhave the following technical features:

[0015] The step of smoothing the amplitude spectrum of the stacked seismic record average trace data to form a zero-phase initial waveletapplicationinclude the following steps:

[0016] The attributes of the triple unit include data attributes and object attributes.

[0017] The step of respectively labeling the head entity array and the tail entity array in the triple unitapplicationinclude the following steps:

[0018] The head start position and the head end position of the head entity array are respectively labeled with a first mark, and the characters between the head start position and the head end position are labeled with a second mark, wherein the first mark is different from the second mark;

[0019] The tail start position and the tail end position of the tail entity array are respectively labeled with a third mark, and the characters between the tail start position and the tail end position are labeled with a fourth mark, wherein the third mark is different from the fourth mark.

[0020] The step of constructing a joint extraction framework model based on the head entity array, the tail entity array and the relationship between the head entity array and the tail entity arrayapplicationinclude the following steps:

[0021] For each character vector in the input text, the head start position and the head end position are calculated respectively, and the calculation formula is:

[0022]

[0023]

[0024] Wherein, c iwherein, for each head entity array in the text, a head start position and a head end position are calculated according to the following formulae: and respectively represent possible positions of the head start position and possible positions of the head end position, and σ is a sigmoid function, and W start and W end respectively represent a start training weight and an end training weight, and b start and b end respectively represent a start training bias and an end training bias.

[0025] The modeling method further comprises the following steps:

[0026] Each of the head entity arrays is mapped to each of the relation-specific taggers, and a tail start position and a tail end position of a tail entity array of each relation are calculated according to the following formulae:

[0027]

[0028]

[0029] wherein, r is a relation type, is an i-th character vector representation of a k-th head entity feature vector and and respectively represent possible positions of the tail start position and possible positions of the tail end position, and pos i represents a part-of-speech vector of a word in which the i-th character is located.

[0030] The method for establishing the knowledge extraction framework based on the knowledge graph using the pre-trained language representation model specifically comprises the following steps:

[0031] The Jieba word segmentation tool is used to perform part-of-speech tagging and embed a part-of-speech vector, and the head entity character vector and the character sequence vector containing sentence information are fused to obtain a vector of a character different from the head entity character position, according to the following formulae:

[0032]

[0033] wherein, c i represents an encoded character vector of the pre-trained language representation model of the i-th character.

[0034] The method for converting the resource description framework in the triple unit into an attribute graph and storing the attribute graph in a Neo4j graph database specifically comprises the following steps:

[0035] The Jena application programming interface is used to read and reason the text, and the Neo4j graph database is used as a storage tool of the attribute graph.

[0036] The converting the resource description framework in the triple unit into an attribute graph and storing the attribute graph in a Neo4j graph database specifically comprises the following steps:

[0037] Extracting triples;

[0038] Reading the ontology model using the Jena application interface;

[0039] Obtaining entity concept information, traversing the triples, and finding the head entity concept and tail entity concept corresponding to the triple relationship in the triples in the ontology model;

[0040] Obtaining entity attribute information, and finding the corresponding attribute name and attribute type in the ontology model according to the head entity concept and the tail entity concept;

[0041] Generating a password statement and storing the triples.

[0042] The application also provides a knowledge graph-based ontology modeling system for flower diseases and pests.

[0043] Additional aspects and advantages of the application will be described in part below, will become apparent from the following description, or will be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0044] The above and / or additional aspects and advantages of the application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which:

[0045] Figure 1 is a framework structure diagram of a knowledge graph-based ontology modeling system for flower diseases and pests according to an embodiment of the application;

[0046] Figure 2 is an ontology model diagram in the field of flower diseases and pests according to an embodiment of the application;

[0047] Figure 3 is a labeling strategy diagram according to an embodiment of the application;

[0048] Figure 4 is a joint extraction framework model diagram according to an embodiment of the application;

[0049] Figure 5 is a RDF2PG mapping algorithm flowchart according to an embodiment of the application. DETAILED DESCRIPTION

[0050] Embodiments of the present application are described below in the context of example embodiments illustrated in the drawings, in which like or similar designations refer to like or similar elements or elements with the same or similar function throughout the several views. The embodiments described below are exemplary in nature and are intended to be illustrative of the present application, and are not to be construed as limiting the present application.

[0051] The knowledge graph is a method for effectively expressing the relationship between data through a semantic network proposed by Google in 2012, and the use of the knowledge graph to manage domain knowledge that is not compatible with traditional knowledge management methods is a research hotspot in various fields.

[0052] Figure 1 is the framework structure diagram of the ontology modeling system of the flower disease and pest based on the knowledge graph of an embodiment of the present application; Figure 2 is the ontology model diagram of the flower disease and pest field of an embodiment of the present application; Figure 3 is the annotation strategy schematic diagram of an embodiment of the present application; Figure 4 is the joint extraction framework model diagram of an embodiment of the present application; Figure 5 is the RDF2PG mapping algorithm flowchart of an embodiment of the present application. Reference Figures 1-5 The present application provides an ontology modeling method of flower disease and pest based on a knowledge graph, which comprises the following steps:

[0053] S1, extracting a plurality of attribute elements of the flower disease and pest field from the text.

[0054] Specifically, ten types of attribute elements of the flower disease and pest field are extracted as main concepts, including flower name, flower growth stage, plant organ, region, disease and pest, chemical fertilizer and pesticide, control method, damage symptom, environment, and pathogen, etc. Key elements for ontology concept attribute extraction.

[0055] S2, constructing an ontology model of the flower disease and pest field, wherein the ontology model comprises a triple unit.

[0056] Specifically, the protégé ontology modeling tool is used to construct the ontology model of the flower disease and pest field, wherein the relationship between each concept is, for example, Figure 1The flower plant disease and pest field ontology model reflects the relationship between the concepts in the flower plant disease and pest field, wherein the subclass is indicated by subClassOf, for example, a (disease, rdfs: subClassOf, plant disease and pest) triple indicates that disease is a subclass of plant disease and pest. The field attributes include data attributes and object attributes, which are indicated by DatatypeProperty and ObjectProperty, respectively. DatatypeProperty indicates the object and type of numerical value attribute, for example, (disease, disease name, xsd: string) indicates that the domain of the attribute disease name of disease is the "disease" class, and the value domain is the string type. ObjectProperty indicates the relationship attribute between classes, for example, (plant disease and pest, damage site, plant organ) indicates that the definition domain and value domain of the relationship "damage site" are the "plant disease and pest" class and the "plant organ" class, respectively. The defined relationship and attribute are constraints on instance data and play a standardizing role on instance data.

[0057] S3, respectively, mark the head entity array and the tail entity array in the triple unit.

[0058] Specifically, the first mark is marked on the head start position and the head end position of the head entity array, and the second mark is marked on the characters between the head start position and the head end position; the first mark is different from the second mark;

[0059] The third mark is marked on the tail start position and the tail end position of the tail entity array, and the fourth mark is marked on the characters between the tail start position and the tail end position; the third mark is different from the fourth mark.

[0060] In a specific implementation, the head-tail entity separation annotation plus the "01" annotation method is used to annotate the triplets. The specific annotation strategy is as follows: (1) First, the head-tail entity separation annotation method is adopted. The annotation sequence array is divided into a head entity sequence array and a tail entity sequence array. Compared with the traditional single sequence array annotation of head-tail entities, the head-tail entities are divided into two independent arrays and are annotated separately, which solves the problem of nested head-tail entity and overlapping head-tail entity annotation. The flower plant disease and pest corpus text is obtained from the Internet and literature, and the semantic triplets are annotated according to the ontology model constructed in steps S1 and S2. (2) Then, the "01" annotation mode is established. Two arrays, the entity start array and the entity end array, are used to represent the start position and the end position of the entity respectively. First, two arrays with the same length as the input text and all elements as "0" are initialized, and then according to the pre-annotated entity content, the head and tail positions of the entity are marked as "1" in the corresponding arrays. When there are multiple possible entities in a sentence, according to the nearest principle, the part between the "1" in the start array and the "1" in the nearest end array is regarded as an entity. Compared with the traditional "BIO" annotation method, the "01" annotation method only needs to perform binary classification label prediction, without the need to predict multiple label categories, reducing the difficulty of prediction. In addition, the "01" annotation method only annotates the head and tail boundary positions of the entity, reducing the probability of entity error or omission during prediction, and at the same time, it can better represent single word entities without introducing additional annotation symbols for separate annotation to further increase the prediction label category and increase the prediction difficulty. Taking the gardenia leaf spot disease as an example, the annotation strategy is as follows: Figure 3 .

[0061] S4, constructing a joint extraction framework model based on the head entity array, the tail entity array, and the relationship between the head entity array and the tail entity array.

[0062] Specifically, the Cas POS Rel triplet is used to construct a joint extraction framework model, that is, the entity and the relationship between the entities are extracted at the same time. For each character vector c i , the possibility of being the start and end position of the head entity is calculated by formula (1) and (2) and

[0063]

[0064]

[0065] σ is the sigmoid activation function (i.e. S-shaped function), W start and W endrespectively represent the start training weight and the end training weight, b start and b end respectively represent the start training bias and the end training bias.

[0066] Then a mapping is established between each head entity and each relationship-specific tagger, and the likelihood of the start and end positions of each relationship tail entity is calculated by formulas (3) and (4) and wherein r is the relationship type, is the kth head entity feature vector, The ith character vector is obtained by adding word features in a manner of combining part-of-speech features with characters as semantic units, thereby fusing word features to obtain a mixed word and part-of-speech feature vector. Finally, the label corresponding to each character is determined according to the set activation threshold.

[0067]

[0068]

[0069] S5, using a pre-trained language representation model to establish a knowledge extraction framework based on a knowledge graph.

[0070] Specifically, an ALBERT (A Lite Bidirectional Encoder Representation from Transformers; a Lite pre-trained language representation model) pre-training model is used as an encoding layer, text features in a corpus are extracted by the ALBERT pre-training model to obtain a character sequence vector with rich semantic information, the obtained character sequence vector is taken as input, passes through a head entity tagger, and the boundary of the most likely head entity is calculated. The start and end positions are represented by "1" in the start array and the end array, respectively. The jieba word segmentation tool is used for part-of-speech tagging and embedding of part-of-speech vectors, and the head entity character vector and the character sequence vector containing sentence information are fused to obtain the ith character vector, as shown in formula (5).

[0071]

[0072] wherein c i represents the ALBERT encoded character vector of the ith character, pos i represents the part-of-speech vector of the word in which the ith character is located, represents the ith character vector of the kth head entity feature vector. The fused feature vector is input into each relationship-specific tagger for tail entity tagging. See the knowledge extraction framework structure diagram for details.

[0073] S6, convert the resource description framework in the triple unit into an attribute graph, and store the attribute graph in a Neo4j graph database.

[0074] Specifically, the RDF2PG mapping algorithm is used to directly store the established triple into the attribute graph, and a flower disease and pest knowledge model management and storage method is provided. In order to ensure the timeliness of the knowledge and the effectiveness of the knowledge discovery based on the knowledge graph, the knowledge graph needs to be updated in time and the storage needs to be controlled in fine granularity. The RDF2PG (Resource Description Framework to Property Graph) mapping method for directly storing the extracted triple into the attribute graph according to the ontology structure stored in the RDF (Resource Description Framework) graph is provided, the ontology file is read and inferred by using the Jena API, and the Neo4j is used as the attribute graph storage tool.

[0075] The ontology modeling method for the flower disease and pest based on the knowledge graph provides a tool and a method for knowledge extraction, knowledge management and knowledge modeling of the knowledge base related to the flower disease and pest control, and provides a new knowledge discovery and knowledge storage and management mode and method based on the knowledge graph for the knowledge base management of the disease and pest expert system, and provides technical support for background knowledge management and knowledge discovery for the diagnosis expert system and intelligent application of the flower disease and pest control.

[0076] The ontology modeling method for the flower disease and pest based on the knowledge graph is used for the text characteristics in the flower disease and pest field, utilizes the multi-feature representation semantics, can realize the joint extraction of the flower disease and pest field entity and the relationship, reduces the knowledge extraction and refining cost, and helps the knowledge graph to be quickly constructed and timely updated. The knowledge management and storage model is combined with the graph database, the RDF2PG mapping method for directly storing the extracted triple into the attribute graph according to the ontology structure stored in the RDF graph is established, and a new mode and method for the flower disease and pest knowledge management and knowledge storage are provided.

[0077] In the specific implementation, referring to Figure 5 One embodiment of the present application provides an RDF2PG mapping algorithm process, and the specific algorithm process is as follows:

[0078] In step S10, the triple is extracted. The corpus to be extracted is input into CasPOSRel, and the extracted triple T is obtained.

[0079] In step S20, the ontology model is read. The ontology model O is read by using the Jena API.

[0080] In step S30, entity concept information is acquired. The triples T in step S7.1 are traversed. The head entity concept DomainClass and the tail entity concept RangeClass corresponding to each triple relationship ObjectProperty in T are looked up in O.

[0081] In step S40, entity attribute information is acquired. According to the head entity concept DomainClass and the tail entity concept RangeClass obtained in step S7.3, the corresponding attribute name DatatypeProperty and attribute type Range are looked up in O.

[0082] In step S50, Cypher statements are generated to store triples. According to the triples obtained in steps S7.1-S7.4 and the semantic model corresponding to the triples in the ontology model, entity addition Cypher statements MERGE(:Class{datatype:instance value}) and relationship addition Cypher statements CREATE UNIQUE(:DomainClass{datatype:instance value})-[:ObjectProperty]->(:RangeClass{datatype:instance value}) are generated. The data is stored in a Neo4j database, and the storage and management of knowledge are completed. Specific embodiments

[0083] The knowledge graph-based ontology modeling method for flower plant diseases and insect pests will be described below with reference to a specific embodiment.

[0084] In the present embodiment, 721 documents, more than 160 types of flowers, and more than 170 types of pests are collected from the documents of "Flower Plant Disease and Pest Control", "Flower Plant Disease and Pest Control: Colorful", "Flower Plant Disease and Pest Control Album", and Baidu Encyclopedia, and the disease symptoms, environmental conditions, and disease control methods caused by the pests are taken as examples.

[0085] Step S11, extract flower plant disease and pest field elements.

[0086] Ten types of elements are taken as main concepts, including flowers, flower growth stages, plant organs, regions, diseases and pests, fertilizers and pesticides, control methods, damage symptoms, environments, and pathogens.

[0087] Step S12, construct a flower plant disease and pest ontology model.

[0088] Using the protégé ontology modeling tool, the relationships of (pests, environmental conditions, environment), (pests, damage site, plant organs), (pests, occurrence area, area), (pests, required chemical fertilizers and pesticides, chemical fertilizers and pesticides), (pests, damage site color, plant organ color), (pests, damage symptoms, plant traits), (pests, damage site shape, plant organ shape), (pests, control methods, control methods), (pests, damage to flowers, flowers), (pests, occurrence period, flower growth stage), (diseases, alternative names, diseases), (diseases, pathogenic pathogens, pathogens), (pests, alternative names, pests) and the like are constructed, and DatatypeProperty attributes are constructed for each category, such as (diseases, disease name, string), (pathogens, pathogen name, string) and the like.

[0089] Step S13, labeling triples.

[0090] Taking gardenia leaf spot disease as an example, the labeling result can be represented as {“text”:“Gardenia leaf spot disease is caused by Phyllosticta gardeniae and Phyllosticta vivi (fungi) infection.”, “triple_list”:[“Gardenia leaf spot disease”, “disease-causing pathogen”, “leaf spot”]}.

[0091] Step S14, labeling entities in triples.

[0092] Taking gardenia leaf spot disease as an example, the labeling strategy is as shown in Figure 3 .

[0093] Step S15, constructing Cas POS Rel extraction framework.

[0094] Taking gardenia leaf spot disease as an example, the model framework is as shown in Figure 4 .

[0095] Step S16, managing and storing knowledge.

[0096] Step S6.1, extracting triples.

[0097] Taking the corpus “Gardenia leaf spot disease is caused by Phyllosticta gardeniae and Phyllosticta vivi (fungi) infection.” as an example, through the extraction framework constructed in step S5, the triples (“Gardenia leaf spot disease”, “disease-causing pathogen”, “leaf spot”) are extracted.

[0098] Step S6.2, reading the ontology model.

[0099] Using Jena API to read the ontology model O constructed in step S2.

[0100] Step S6.3, obtaining entity concept information.

[0101] In O, find the head entity concept DomainClass "disease" and the tail entity concept RangeClass "pathogen" corresponding to the "disease pathogen" relationship.

[0102] Step S6.4, obtain entity attribute information.

[0103] Find the DatatypeProperty corresponding to "disease" and "pathogen" respectively, and get "disease name" and "pathogen name" and the value domain of the two attributes, which are both string.

[0104] Step S6.5, generate Cypher statements to store triples.

[0105] Generate entity addition Cypher statements MERGE(:disease{disease name: 'gardenia leaf spot disease'}), MERGE(:pathogen{pathogen name: 'leaf spot'}) and relationship addition Cypher statements CREATE UNIQUE(:disease{disease name: 'gardenia leaf spot disease'})-[:disease pathogen]->(:pathogen{pathogen name: 'leaf spot'}) to complete the storage of the triple ("gardenia leaf spot disease", "disease pathogen", "leaf spot").

[0106] The application also provides a knowledge graph-based ontology modeling system for flower diseases and pests, which is realized by any of the above knowledge graph-based ontology modeling methods for flower diseases and pests.

[0107] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0108] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments without departing from the principles and spirit of the present application within the scope of the present application.

Claims

1. A method for ontology modeling of flower diseases and pests based on a knowledge graph, characterized in that, The method comprises the following steps: extracting multiple attribute elements in the field of flower diseases and insect pests from the text; constructing an ontology model in the field of flower diseases and insect pests, wherein the ontology model comprises a triple unit; annotating a head entity array and a tail entity array in the triple unit respectively; constructing a joint extraction framework model based on the head entity array, the tail entity array and the relationship between the head entity array and the tail entity array; establishing a knowledge extraction framework based on a knowledge graph using a pre-trained language representation model; converting a resource description framework in the triple unit into an attribute graph, and storing the attribute graph in a Neo4j graph database; The annotation of the head entity array and the tail entity array in the triple unit respectively comprises the following steps: annotating a head start position and a head end position of the head entity array with a first mark respectively, and annotating characters between the head start position and the head end position with a second mark, wherein the first mark is different from the second mark; annotating a tail start position and a tail end position of the tail entity array with a third mark respectively, and annotating characters between the tail start position and the tail end position with a fourth mark, wherein the third mark is different from the fourth mark; The construction of the joint extraction framework model based on the head entity array, the tail entity array and the relationship between the head entity array and the tail entity array comprises the following steps: For each character vector in the input text, the head start position and the head end position are calculated respectively, and the calculation formula is: wherein c i is a single character vector in the text, and respectively represent possible positions of the head start position and possible positions of the head end position, σ is a sigmoid function, W start and W end respectively represent the start training weight and the end training weight, b start and b end respectively represent the start training bias and the end training bias. 2.The knowledge graph-based ontology modeling method for flower diseases and pests according to claim 1, characterized in that, The attributes of the triple unit include data attributes and object attributes. 3.The knowledge graph-based ontology modeling method for flower diseases and pests according to claim 1, characterized in that, The modeling method further comprises the following steps: mapping each head entity array with each relationship-specific annotator, and calculating the tail start position and the tail end position of the tail entity array of each relationship, and the calculation formula is: wherein r is a relation type, is the i-th character vector representation of the k-th head entity feature vector is the i-th character vector representation of the k-th head entity feature vector and denote the possible positions of the start of the tail and the end of the tail, respectively, pos i denotes the part-of-speech vector of the word in which the i-th character is located. 4.The method of claim 3, wherein, The establishment of the knowledge extraction framework based on the knowledge graph using the pre-trained language representation model comprises the following steps: performing part-of-speech tagging using the Jieba segmentation tool and embedding part-of-speech vectors, fusing head entity character vectors and character sequence vectors containing sentence information to obtain vectors of characters different from the head entity character position, and the calculation formula is as follows: where c i represents the encoding character vector of the pre-trained language representation model of the i-th character. 5.The knowledge graph-based ontology modeling method for flower diseases and pests according to claim 1, characterized in that, The conversion of the resource description framework in the triple unit into an attribute graph, and the storage of the attribute graph in a Neo4j graph database comprises the following steps: reading and reasoning the text using the Jena application programming interface, and using the Neo4j graph database as a storage tool for the attribute graph. 6.The method of claim 5, wherein the method further comprises: The conversion of the resource description framework in the triple unit into an attribute graph, and the storage of the attribute graph in a Neo4j graph database comprises the following steps: extracting triples; reading the ontology model using the Jena application programming interface; obtaining entity concept information, traversing the triples, and finding the head entity concept and the tail entity concept corresponding to the triple relationship in the triples in the ontology model; Acquire entity attribute information, and find corresponding attribute name and attribute type in the ontology model according to the head entity concept and the tail entity concept; Generate a password statement, and store the triple.

7. An ontology modeling system for flower diseases and pests based on a knowledge graph, characterized by, The modeling system is implemented by the ontology modeling method for flower diseases and insect pests based on a knowledge graph according to any one of claims 1-6.

Images