Knowledge graph completion method for DHI report interpretation, DHI report interpretation method, storage medium and equipment

Abstract

The invention discloses a knowledge graph completion method for DHI report interpretation, a DHI report interpretation method, a storage medium and equipment, and belongs to the technical field of livestock breeding. In order to solve the problem that a knowledge graph used for DHI report interpretation at present is constructed based on known information and cannot effectively interpret potential performance indexes / symptoms and influence factors, so that DHI report interpretation is not comprehensive. The method comprises the following steps: firstly, based on a basic text library, determining entities and entity relationships by utilizing known relationships, and constructing a reference DHI domain knowledge graph in combination with a crowdsourcing calculation mode; according to the method, the knowledge graph is constructed, then the knowledge graph is complemented by predicting the entity relationship based on a large oracle model, meanwhile, the edge weight of the incremental knowledge graph is determined through objective logic relationship representation, the edge weight of the basic knowledge graph is adjusted, and the knowledge graph capable of containing potential performance indexes / symptoms and influence factors can be constructed.

Description

Technical Field

[0001] The present invention belongs to the technical field of livestock breeding, and relates to a method for completing a knowledge graph for DHI report interpretation, a DHI report interpretation method, a storage medium, and a device. Background Art

[0002] DHI (Dairy Herd Improvement) refers to a milk production performance measurement system, which mainly measures the milk production and milk components of dairy cows, collects herd information, and regularly forms a detection and guidance report after analysis. The correct interpretation of the DHI report can effectively guide the feeding, breeding, and management of dairy farms.

[0003] Currently, the interpretation of DHI reports basically depends on experts with professional experience. However, experts with professional experience are not fully equipped in all ranches. Therefore, most ranches are unable to correctly and effectively interpret DHI reports, not only unable to effectively utilize the value of DHI reports, but also unable to give scientific and effective guidance on the feeding management of dairy herds. In addition, due to the differences in professional experience among experts with professional experience, there are also certain differences in the interpretation of DHI reports. Therefore, a unified standard for DHI report interpretation cannot be formed, resulting in inconsistent management systems for different ranches and certain problems in the scientificity and effectiveness of the feeding of dairy herds.

[0004] Gao Meng et al. from Northeast Agricultural University proposed a "method for interpreting DHI reports based on a knowledge graph (application number CN202110969609.X)". This method combines the DHI domain knowledge graph to diagnose problems based on the results of dynamic analysis. Based on the DHI domain knowledge graph, the DHI domain knowledge graph includes three types of entities and entity relationships: "performance indicators / symptoms", "influence factors", and "solutions". Therefore, it realizes the automatic interpretation of DHI reports, enabling any dairy herd feeding management unit to obtain scientific and effective guidance; at the same time, a unified standard for DHI report interpretation is formed, enabling different ranch managements to form a unified system, ensuring the scientificity and effectiveness of the feeding of dairy herds. However, its effect Summary of the Invention

[0005] The present invention aims to solve the problem that the knowledge graphs currently used for DHI report interpretation are constructed based on known information and cannot effectively interpret potential performance indicators / symptoms and influence factors, resulting in incomplete DHI report interpretation.

[0006] A method for completing a knowledge graph for DHI report interpretation includes the following steps:

[0007] First, obtain the benchmark DHI domain knowledge graph, and record the weight of the edges in the benchmark DHI domain knowledge graph as the basic edge weight ω. m1 ; Then, complete the benchmark DHI domain knowledge graph, including the following steps:

[0008] S201. Take all entities in the benchmark DHI domain knowledge graph as reference entities, extract entities and entity relationships in all texts in the incremental text library based on the language large model, and predict the tail entity through the language large model to obtain the predicted triple (h, r, t), where h and t are used to represent the head entity and the tail entity in the triple, and r is used to represent the head entity relationship in the triple; Based on the predicted triple, complete the structure of the benchmark DHI domain knowledge graph to obtain the complete DHI domain knowledge graph structure, denoted as the full DHI domain knowledge graph, and the difference part between the full DHI domain knowledge graph and the benchmark DHI domain knowledge graph is denoted as the incremental DHI domain knowledge graph;

[0009] The neighbor nodes of a certain node in the benchmark DHI domain knowledge graph are denoted as basic connection nodes, and the corresponding edges are basic edges. The neighbor nodes in the incremental DHI domain knowledge graph are incremental connection nodes, and the corresponding edges are incremental edges; For a certain node, calculate the ratio k of the number of its corresponding incremental edges to the total number of edges m , as the incremental adjustment coefficient; Take 1 - k m as the basic adjustment coefficient;

[0010] S202. Based on the two types of entities "influence factors" and "performance indicators / symptoms" in the DHI knowledge graph, count the non-repeated triples (h, r, t) corresponding to all entities in the incremental text library; Among all non-repeated triples (h, r, t), for a certain entity v m , denote the triples containing entity v m as Count the quantity n m , and then based on n m obtain the incremental edge weight ω m corresponding to entity v m2 ;

[0011] S203. Modify the basic edge weight in the construction process of the benchmark DHI domain knowledge graph to (1 - k m )ω m1 , and modify the incremental edge weight to k m ω m2 ; Thus, complete the complement of the benchmark DHI domain knowledge graph.

[0012] Furthermore, the construction process of the benchmark DHI domain knowledge graph includes:

[0013] (1)Construct the DHI domain ontology, which includes three types of entities and entity relationships: "performance indicators / symptoms", "influencing factors", and "solutions"; the performance indicators / symptoms refer to the indicators used to present the health status of dairy cows or the symptoms exhibited by dairy cows;

[0014] (2)Use the electronic text obtained by digitizing the DHI measurement and application guidance related materials as the annotation object, and use the ontology as the annotation basis to perform semantic annotation on the electronic text data to form annotation data;

[0015] (3)Use the data in the annotation data as training data, and extract entities and entity relationships from the text in the basic text library according to the ontology structure of the DHI domain knowledge graph to obtain entity and entity relationship data, construct triples of any two types of entities and entity relationships, and the DHI domain knowledge graph, denoted as the benchmark DHI domain knowledge graph;

[0016] (4)For the triples in the DHI knowledge graph that contain two types of entities, "influencing factors" and "performance indicators / symptoms", calculate the conditional probability between the two types of entities, denoted as P(fac|sym), as the weight of the edge between entities, denoted as the basic edge weight ω m1 。

[0017] Furthermore, the P(fac|sym) is obtained through a crowdsourcing calculation method.

[0018] Furthermore, based on n m obtain the entity v m corresponding incremental edge weight ω m2 The process includes:

[0019] For a certain entity v m , count the number of documents in the incremental text library that contain and calculate the incremental edge weight corresponding to the entity v where N is the number of documents in the incremental text library. m where N is the number of documents in the incremental text library.

[0020] Furthermore, the knowledge graph uses the CompGCN network as the network framework;

[0021] Furthermore, the language large model selects the BERT model.

[0022] A method for interpreting DHI reports includes:

[0023] First, obtain the DHI data of the ranch;

[0024] Then, use the knowledge graph completion method for DHI report interpretation to interpret the DHI report using the completed knowledge graph.

[0025] A computer storage medium stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement the DHI report interpretation method.

[0026] A DHI report interpretation device includes a processor and a memory. The memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the DHI report interpretation method.

[0027] Beneficial effects:

[0028] Based on the large prediction model to predict entity relationships to complete the knowledge graph, and at the same time determine the edge weights of the incremental knowledge graph through the objectively existing logical relationship representations and adjust the edge weights of the basic knowledge graph. The present invention can construct a knowledge graph that can include potential performance indicators / symptoms and influencing factors, and can realize the interpretation of potential performance indicators / symptoms and influencing factors. Therefore, the DHI report interpretation is more comprehensive and effective. Description of the drawings

[0029] Figure 1 It is a schematic diagram of the DHI domain ontology.

[0030] Figure 2 It is a schematic diagram of the knowledge graph completion process.

[0031] Figure 3 It is a schematic diagram of the report interpretation process based on the completed knowledge graph. Detailed implementation manners Detailed implementation manner one:

[0033] This implementation manner is a DHI report interpretation method, which is actually a DHI report interpretation method based on knowledge graph completion. It includes two parts. One part is the knowledge graph completion method for DHI report interpretation, and the other part is the DHI report interpretation method using the completed knowledge graph.

[0034] First, complete the knowledge graph for DHI report interpretation, including the following steps:

[0035] S1. Construction of the benchmark DHI domain knowledge graph:

[0036] (1) Construct the DHI domain ontology. The DHI domain ontology includes three types of entities and entity relationships: "performance indicators / symptoms", "influencing factors", and "solution measures", as Figure 1 shown; the performance indicators / symptoms refer to the indicators used to present the health status reflected by cows or the symptoms reflected by cows;

[0037] (2) Use the electronic text obtained by digitizing DHI reports and professional books, literature, etc. related to application guidance as the annotation object, and take the ontology as the annotation basis to perform semantic annotation on the electronic text data to form annotation data;

[0038] (3) Use the data in the annotation data as training data, and according to the ontology structure of the DHI domain knowledge graph, adopt supervised, semi-supervised, and unsupervised methods to extract entities and entity relationships from the texts in the basic text library to obtain entity and entity relationship data, construct triples of any two types of entities and entity relationships, and the DHI domain knowledge graph, denoted as the benchmark DHI domain knowledge graph; The knowledge graph uses the CompGCN network as the network framework. The basic text library includes professional books, literature, DHI reports, etc. on dairy cows and pastures, as well as texts on the Internet such as Baidu Encyclopedia, which can be constructed according to the actual situation. It should be noted that the more texts in this basic text library, the better the effect of constructing the benchmark DHI domain knowledge graph in theory, but the workload will increase greatly. The present invention can appropriately construct the size of the basic text library, but it needs to be as comprehensive as possible to cover all known entities and entity relationships.

[0039] (4) For the triples in the DHI knowledge graph that contain two types of entities, "influence factors" and "performance indicators / symptoms", calculate the conditional probability between the two types of entities, denoted as P(fac|sym), as the weight of the edge between the entities, denoted as the basic edge weight ω m1 。

[0040] P(fac|sym) is obtained through the crowdsourcing calculation method, that is: through the crowdsourcing software, all the influence factors that may cause a certain performance indicator / symptom are provided to the participants (pasture production personnel, pasture management personnel, domain experts, etc.), and each participant ranks and scores the influence degree of these factors; According to the data, calculate the weights of each influence factor and a certain performance indicator / symptom according to the principle of ranking first and then scoring, that is, follow the principle of the minority obeying the majority, select the influence factor with the most occurrences in each position as the influence factor in that position, and calculate the average value of the scoring values of the participants of the influence factor in that position as the weight between the influence factor and the performance indicator / symptom. Use the same method to obtain the correlation strength between each performance indicator / symptom and each influence factor, and form a correlation coefficient matrix, which represents the weight of the edge between the two types of entities, "performance indicator / symptom" and "influence factor".

[0041] S2. Completion of the benchmark DHI domain knowledge graph:

[0042] For existing professional books, literature, DHI reports, etc. on dairy cows and pastures, the annotation by marking data is based on the entities and the relationships between entities determined by the recognized knowledge. For example, only when it is already known which "influence factors" will affect a certain "performance indicator / symptom", will the entities and the entity relationships be marked and extracted. And the extraction of entity and entity relationships based on supervised, semi-supervised, or even unsupervised methods is basically based on this "known" information. However, there will be a situation where "objectively, there are actually relationships between entities, but we don't know them yet". This situation will lead to the problem that the constructed knowledge graph is actually incomplete, thus restricting the interpretation effect of the knowledge graph to a certain extent. Therefore, this embodiment focuses on completing the knowledge graph in the benchmark DHI field:

[0043] S201. Take all entities in the benchmark DHI field knowledge graph as reference entities, extract the entities and entity relationships (h, r,?) in all texts in the incremental text library based on the BERT model, and predict the tail entity through the BERT model to obtain the predicted triple (h, r, t);

[0044] Based on the predicted triples, complete the structure of the benchmark DHI field knowledge graph to obtain the complete DHI field knowledge graph structure, denoted as the full-scale DHI field knowledge graph. The difference part between the full-scale DHI field knowledge graph and the benchmark DHI field knowledge graph is denoted as the incremental DHI field knowledge graph;

[0045] The neighbor nodes of a certain node in the benchmark DHI field knowledge graph are denoted as basic connection nodes, and the corresponding edges are basic edges. The neighbor nodes in the incremental DHI field knowledge graph are incremental connection nodes (actually the complemented nodes), and the corresponding edges are incremental edges. For a certain node, calculate the ratio k of the number of its corresponding incremental edges to the total number of edges m , as the incremental adjustment coefficient; Take 1 - k m as the basic adjustment coefficient. For example, a certain node has four neighbor nodes in the benchmark DHI field knowledge graph. After completion, there are a total of six nodes, of which two newly added are incremental connection nodes. The edges corresponding to the node and the incremental connection nodes are incremental edges. For this node, the incremental adjustment coefficient The basic adjustment coefficient is

[0046] S202. Based on the two types of entities "influence factors" and "performance indicator / symptom" included in the DHI knowledge graph, count the non-repeated triples (h, r, t) corresponding to all entities in the incremental text library; Among all non-repeated triples (h, r, t), for a certain entity v m , the triples containing the entity v m are denoted as Statistics The quantity n of m During this process, the entity v m should be counted for the quantity of triples corresponding to the head entity h and the tail entity t;

[0047] For a certain entity v m , count the number of documents in the incremental text library that contain and calculate the incremental edge weight corresponding to the entity v where N is the number of documents in the incremental text library. m

[0048] During the process of complementing the reference DHI domain knowledge graph, essentially, it is necessary to supplement the entities and the relationships between them that objectively exist but are not yet known. Also, precisely because the relationships between entities are not yet known to us, the edge weights cannot be obtained through step (4). To determine the edge weights that cannot be obtained through (4), the present invention turns to mining the objectively existing conditions. Since the texts already existing in the incremental text library objectively record the relationships between entities, although some relationships are not known, they objectively exist in reality. Therefore, based on the BERT model, these unknown but actually existing relationships in the incremental text library can be extracted, and the text quantity corresponding to a certain entity v in the incremental text library m can, to a certain extent, characterize the strength of the coexistence connection between the entity and the relationships objectively. Therefore, as long as the text quantity in the incremental text library is ensured to be sufficient, the representation of the relationships can be achieved. So, the present invention calculates the incremental edge weights by using the objectively existing phenomena. During the process of calculating the edge weights, considering that if the probability of a certain objectively potential relationship holding is higher, then its occurrence probability should also be higher, or rather, its potential "manifestation" should be more frequent. Therefore, the present invention uses the document appearance frequency (the proportion of the entity document quantity) corresponding to a certain entity relationship as its objectively potential probability. At the same time, considering that for the potential relationships extracted by the language large model, this kind of potential relationship is essentially an "estimation result", that is, a certain entity is estimated to have an entity relationship with other entities. When a certain entity has more relationships with other entities, it indicates that its representation ability is weaker, and therefore its relationship confidence level is relatively lower. For example, a certain entity v A has entity relationships with ten entities v a1 -v a10 This indicates that there may be ten influencing factors that can all lead to the appearance of a certain 1 symptom or the abnormality of a certain 1 indicator; if the entity v A only has entity relationships with 3 entities v a1 -v a3 ​​There are entity relationships, indicating that only three influencing factors can cause an abnormality in a certain symptom or a certain indicator. It can be seen that for the estimated potential relationships, compared with "one symptom / one indicator is related to ten influencing factors", "one symptom / one indicator is related to three influencing factors" indicates that the representational ability of the symptom / indicator is stronger and its confidence level is higher. Therefore, the present invention uses (h, r, t) vm The quantity n m The reciprocal represents the confidence level and participates in the calculation of the edge weight.

[0049] S203. Modify the basic edge weight in the process of constructing the benchmark DHI domain knowledge graph to (1 - k m )ω m1 , and modify the incremental edge weight to k m ω m2 ;

[0050] So far, the completion of the benchmark DHI domain knowledge graph is completed. The process of knowledge graph completion is as Figure 2 shown. Then, use the completed DHI domain knowledge graph to interpret the DHI report. The report interpretation process can be carried out according to the DHI report interpretation method based on the knowledge graph in the prior art. The interpretation process is as Figure 3 shown. In some embodiments, the process of interpretation includes:

[0051] S301. Acquisition of DHI index data:

[0052] Obtain the DHI data of the ranch. The DHI index data includes two types: this month's data and historical data. This month's data is directly obtained by automatically extracting the DHI report file made by the DHI testing center according to the China Dairy Cattle Performance Testing and Analysis System (CNDHI) through software. This method is simple and efficient and can provide basic measurement data and related statistical indicators, such as average calving interval, lactation days, milk fat rate, protein rate, fat-protein ratio, peak milk, peak day, persistency, urea nitrogen, etc.; historical data is obtained by the software traversing the relevant index data in the historical DHI report file and storing the relevant index data into the database according to the established fields.

[0053] S302. Statistical analysis of index data:

[0054] The analysis of index data includes two aspects: static analysis and dynamic analysis;

[0055] Static analysis is to find abnormal indicators according to this month's data of each indicator according to the normal range value standard of each indicator and form a corresponding factual description;

[0056] Dynamic analysis is to analyze the recent change rules of each indicator by combining this month's data and historical data of each indicator and form a corresponding factual description.

[0057] S303. Problem diagnosis:

[0058] Combined with the DHI domain knowledge graph, problem diagnosis is performed on the results of dynamic analysis. Problem diagnosis includes two aspects: problem location and guidance on measures.

[0059] Problem location is based on the DHI domain knowledge graph. Taking the factual description of dynamic analysis as the "performance indicator / symptom" entity, the probability that the factual description is affected by a certain influencing factor is calculated, denoted as P(fac):

[0060] P(fac) = P(fac|sym)·P prior (sym)

[0061] where P(fac|sym) is the conditional probability between the performance indicator / symptom and the influencing factor, that is, the weight of the edge between entities (the modified basic edge weight and the incremental edge weight); P prior (sym) is the prior probability of the performance indicator / symptom. P prior (sym) The initial value is calculated based on historical DHI reports and ranch record data, that is: the prior probability of the performance indicator is the proportion of the number of reports with abnormal indicators in the historical DHI reports to the total number of reports; the prior probability of the symptom is the proportion of the number of cows with the symptom in the ranch record data to the total number of cows. In addition, P prior (sym) can be updated monthly;

[0062] The guidance on measures is to find the corresponding solution measures for the influencing factor according to the relationship between the two types of entities "influencing factor" and "solution measure" from the DHI domain knowledge graph based on the located influencing factor, and feedback it to the user. Specific implementation method two:

[0064] For the method of interpreting DHI reports using the completed DHI domain knowledge graph, the corresponding processing process is coded to obtain the corresponding interpretation program; then it is stored in a computer storage medium of this implementation method, that is, there is at least one instruction stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the described method of interpreting DHI reports.

[0065] It should be understood that the instructions include a computer program product, software, or computerized method corresponding to any method described in the present invention; the instructions can be used to program a computer system or other electronic devices. A computer storage medium can include a readable medium on which instructions are stored, which can include but are not limited to magnetic storage media and optical storage media; magneto-optical storage media include read-only memory ROM, random access memory RAM, erasable programmable memory (e.g., EPROM and EEPROM), and flash memory layers, or other types of media suitable for storing electronic instructions. Specific Embodiment 3:

[0067] For the method of interpreting a DHI report using the completed DHI domain knowledge graph, the corresponding processing process is coded to obtain a corresponding interpretation program; then it is stored in a memory and run on a DHI report interpretation device described in this embodiment, that is, the device includes a processor and a memory. It should be understood that it includes any device including a processor and a memory described in the present invention. The device may further include other units and modules for display, interaction, processing, control, etc. through signals or instructions and other functions.

[0068] 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 described method for interpreting a DHI report.

[0069] Those skilled in the art should understand that the stored at least one instruction is a computer program product corresponding to the method or system. Therefore, the present application can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code. The solutions in the embodiments of the present application can be implemented using various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript.

[0070] The present application is described with reference to the flowcharts and / or block diagrams of methods, systems, and computer program products according to embodiments of the present application, and can also be used for corresponding devices. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocksFigure 1 means for the functions specified in one or more blocks.

[0071] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one Figure 1 process or more processes and / or blocks Figure 1 means for the functions specified in one or more blocks.

[0072] These computer program instructions may also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 process or more processes and / or blocks Figure 1 means for the functions specified in one or more blocks.

[0073] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications falling within the scope of the present application.

[0074] Obviously, those skilled in the art can make various changes and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

[0075] The present invention may also have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and deformations according to the present invention, but these corresponding changes and deformations should all fall within the protection scope of the claims appended to the present invention.

Claims

1. A knowledge graph completion method for DHI report interpretation, characterized in that Including the following steps: First, obtain the benchmark DHI domain knowledge graph, and denote the weight of the edges in the benchmark DHI domain knowledge graph as the basic edge weight ω m1 ; then, complete the benchmark DHI domain knowledge graph, including the following steps: S201: Take all entities in the benchmark DHI domain knowledge graph as reference entities, extract entities and entity relationships in all texts in the incremental text library based on the language model, and predict the tail entity through the language model to obtain the predicted triples (h, r, t), where h and t are used to represent the head entity and the tail entity in the triple, and r is used to represent the head entity relationship in the triple; complete the structure of the benchmark DHI domain knowledge graph based on the predicted triples to obtain the complete DHI domain knowledge graph structure, denoted as the full DHI domain knowledge graph, and the difference between the full DHI domain knowledge graph and the benchmark DHI domain knowledge graph is denoted as the incremental DHI domain knowledge graph; The neighbor nodes of a certain node in the baseline DHI domain knowledge graph are denoted as basic connection nodes, and the corresponding edges are basic edges. The neighbor nodes in the incremental DHI domain knowledge graph are incremental connection nodes, and the corresponding edges are incremental edges. For a certain node, calculate the ratio k of the number of its corresponding incremental edges to the total number of edges. m , as the incremental adjustment coefficient; take 1 - k m as the basic adjustment coefficient. S202. Based on the two types of entities, "influence factors" and "performance indicators / symptoms", in the DHI knowledge graph, count the non-repeated triples (h, r, t) corresponding to all entities in the incremental text library; among all the non-repeated triples (h, r, t), for a certain entity v m , the triples containing the entity v m are denoted as count the quantity n m , and then based on n m obtain the incremental edge weight ω m corresponding to the entity v m2 ; S203. Modify the basic edge weight in the construction process of the benchmark DHI domain knowledge graph to (1 - k m )ω m1 , and modify the incremental edge weight to k m ω m2 ; thus completing the completion of the benchmark DHI domain knowledge graph.

2. The knowledge graph completion method for DHI report interpretation according to claim 1, wherein The construction process of the benchmark DHI domain knowledge graph includes: (1) Construct a DHI domain ontology, which includes three types of entities and entity relationships: "performance indicators / symptoms", "influence factors", and "solutions"; the performance indicators / symptoms refer to the indicators used to present the health status reflected by dairy cows or the symptoms shown by dairy cows; (2) Use the electronic text obtained by digitizing the DHI measurement and application guidance-related materials as the annotation object, and perform semantic annotation on the electronic text data based on the ontology to form annotation data; (3) Use the data in the annotation data as training data, and extract entities and entity relationships from the texts in the basic text library according to the ontology structure of the DHI domain knowledge graph to obtain entity and entity relationship data, construct triples of any two types of entities and entity relationships, and the DHI domain knowledge graph, denoted as the benchmark DHI domain knowledge graph; (4) For the triples in the DHI knowledge graph that contain two types of entities, namely "influence factors" and "performance indicators / symptoms", calculate the conditional probability between the two types of entities, denoted as P(fac|sym), and use it as the weight of the edge between the entities, denoted as the basic edge weight ω m1 .

3. The knowledge graph completion method for DHI report interpretation according to claim 2, wherein The P(fac|sym) is obtained through a crowdsourcing calculation method.

4. The knowledge graph completion method for DHI report interpretation according to any one of claims 1 to 3, characterized in that Based on n m obtain entity v m corresponding incremental edge weight ω m2 The process includes: For an entity v m , the statistical incremental text library contains Number of documents And calculate the entity v m The corresponding incremental edge weight Where N is the number of documents in the incremental text library.

5. The knowledge graph completion method for DHI report interpretation according to claim 4, wherein The knowledge graph uses the CompGCN network as the network framework.

6. The knowledge graph completion method for DHI report interpretation according to claim 5, wherein The language model selects the BERT model.

7. A method for interpreting DHI reports, characterized in that, Including: First, obtain the DHI data of the ranch; Then, use the knowledge graph completed by the knowledge graph completion method for DHI report interpretation described in any one of claims 1 to 6 to interpret the DHI report.

8. A computer storage medium, characterized in that The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement a DHI report interpretation method according to claim 7.

9. A DHI report interpretation device, characterized in that, The device includes a processor and a memory, and the memory stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement a DHI report interpretation method according to claim 7.

Images