Knowledge graph-based biological asset supervision method and computing device

Through a knowledge graph-based method, combined with text and image data, automatic identification and full-time supervision of beef cattle biological assets are realized, and the problems of high labor intensity and incomplete monitoring of traditional regulatory methods are solved, and the accuracy and efficiency of supervision are improved.

CN118982436AInactive Publication Date: 2024-11-19北京牧国科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410944786.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The problems of high labor intensity and incomplete monitoring of traditional beef cattle biological assets supervision methods have made it difficult to detect and deal with diseases and behavior abnormalities in a timely manner, and the lack of accurate data support has limited the improvement of agricultural production efficiency.

Method used

Using a knowledge graph-based method, we use the knowledge graph to construct a knowledge graph from text data, combine image acquisition and behavior recognition, and build query vectors for reasoning, so as to realize automatic identification of biological assets and full-time supervision.

Benefits of technology

It reduces labor intensity, achieves efficient and accurate supervision of biological assets, improves the accuracy of disease prediction and behavioral analysis, and provides scientific support for agricultural production.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982436A_ABST
    Figure CN118982436A_ABST
Patent Text Reader

Abstract

The invention provides a knowledge graph-based biological asset supervision method and computing equipment, and the method comprises the steps: pre-constructing a knowledge graph from text data describing the experience knowledge of biological assets, and complementing the knowledge graph through knowledge graph embedding; carrying out image acquisition on the biological assets; performing basic behavior identification on the biological assets according to the collected images; constructing a query vector by using the identified behavior tag; and utilizing the query vector to execute reasoning operation on the knowledge graph, converting a reasoning result into a recognizable form, and outputting the recognizable form as a supervision result. According to the method, the extensive management mode of the breeding industry can be improved, and the accuracy of biological asset supervision is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and in particular to a method and computing device for biological asset supervision based on knowledge graph. Background Art

[0002] With the development of the beef cattle industry and the need to improve agricultural and animal husbandry production efficiency, higher requirements have been placed on the biological asset supervision of the beef cattle asset governance system. Traditional beef cattle breeding is a labor-intensive industry, and it is urgently needed to upgrade the industry to improve production efficiency. Since beef cattle cannot directly express their feelings, continuous observation and judgment are required to achieve biological asset supervision. The identification of the growth process of animals and plants through machine learning models has achieved beneficial effects in biological asset supervision. For example, for disease prediction of biological assets, estrus monitoring and precision feeding monitoring, the application of machine learning models to identify the growth process of animals and plants has promoted improvements in many fields such as disease prediction, estrus regulation, and precision feeding management.

[0003] Biological asset management is considered an effective solution to improve agricultural production efficiency and promote agricultural science and technology. The current model for beef cattle supervision lacks experience and knowledge and is not very accurate. In addition, the supervision of each indicator requires a specially trained model to complete the task, and the intelligent process of beef cattle biological asset supervision is slow.

[0004] To this end, a technical solution is needed to address the problems of high labor intensity and incomplete monitoring of traditional beef cattle biological asset supervision methods. Summary of the invention

[0005] The present invention aims to provide a method for biological asset supervision based on knowledge graph, which can realize automatic identification of biological assets, reduce labor intensity, and provide all-round supervision at all times.

[0006] According to one aspect of the present invention, there is provided a method for biological asset supervision based on a knowledge graph, comprising pre-constructing a knowledge graph from text data describing empirical knowledge of biological assets, and completing the knowledge graph by embedding the knowledge graph;

[0007] Capturing images of the biological assets;

[0008] Performing basic behavior recognition on the biological asset based on the collected images;

[0009] Constructing a query vector using the identified behavior labels;

[0010] The query vector is used to perform reasoning operations in the knowledge graph, and the reasoning results are converted into a recognizable form and output as supervision results.

[0011] According to some embodiments, pre-building a knowledge graph from text data describing empirical knowledge of biological assets and completing the knowledge graph by embedding the knowledge graph includes:

[0012] The text data describing the experience knowledge of biological assets is constructed into word sequences, each word forms a corresponding word embedding vector, and the word embedding vector is input into the encoder for feature extraction;

[0013] Passing the feature vector extracted by the encoder to a linear classifier for entity recognition;

[0014] Extract relationships from identified entities to obtain a knowledge graph that includes entities and relationships;

[0015] The knowledge graph embedding method is used to discover the implicit relationships between entities in the knowledge graph, and the knowledge graph is completed based on the implicit relationships.

[0016] According to some embodiments, the encoder comprises: a Transformer encoder based on a Bert model.

[0017] According to some embodiments, capturing images of the biological asset includes:

[0018] Obtain video image data of the daily behavior activities of the biological asset.

[0019] According to some embodiments, performing basic behavior recognition on the biological asset based on the collected image includes:

[0020] A convolutional neural network is used to identify the behavior of the biological asset in the video image data.

[0021] According to some embodiments, extracting relationships from identified entities to obtain a knowledge graph including entities and relationships includes:

[0022] The entities and the relationships are constructed into triples as basic units for constructing the knowledge graph.

[0023] According to some embodiments, the textual data includes: veterinary reports, agricultural research papers, and / or farm daily records.

[0024] According to some embodiments, the knowledge graph includes a knowledge graph of the correspondence between cattle behaviors and diseases.

[0025] According to another aspect of the present invention, a computing device is provided, comprising a processor and a memory storing a computer program, wherein when the computer program is executed by the processor, the processor executes any of the above methods.

[0026] According to another aspect of the present invention, there is provided a non-transitory computer-readable storage medium having computer-readable instructions stored thereon, and when the instructions are executed by a processor, the processor is caused to execute any of the above methods.

[0027] The biological asset supervision method based on knowledge graph in this example embodiment can organically combine textual knowledge and visual observation data, construct a knowledge graph to convert the empirical knowledge of biological assets into a knowledge graph, and supplement and improve it through knowledge graph embedding, making the information more orderly and accessible. By constructing a query vector, the behavioral information of biological assets can be converted into a compact mathematical representation, which not only facilitates storage and transmission, but also improves the efficiency of reasoning and query operations. It realizes intelligent monitoring and analysis of biological asset behavior and provides scientific support for biological resource management and protection.

[0028] The biological asset supervision solution based on knowledge graph in this example embodiment performs behavioral image recognition on the basic actions obtained from image acquisition, and uses the knowledge graph constructed by empirical knowledge to quickly expand to a general framework method for multiple recognition tasks. It realizes automatic recognition of biological assets, reduces labor intensity, and provides all-round supervision at all times.

[0029] This example embodiment realizes the construction method of the knowledge graph by extracting knowledge from text descriptions, and combines various high-tech technologies such as image recognition, natural language processing, and knowledge graphs to achieve the supervision of living biological assets, providing efficient and reliable digital management for the supervision of living animal assets, which is conducive to improving the extensive management model of the animal husbandry industry. By constructing a general supervision framework, the knowledge graph can be quickly used to build expanded biological asset supervision capabilities, thereby improving the accuracy of biological asset supervision.

[0030] It is to be understood that the foregoing general description and the following detailed description are exemplary only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for describing the embodiments are briefly introduced below.

[0032] Figure 1 A schematic diagram of an architecture for building a knowledge graph according to an example embodiment is shown.

[0033] Figure 2 A flowchart of a method for implementing knowledge graph-based biological asset management according to an example is shown.

[0034] Figure 3 A flow chart of a method for objective reasoning for biological asset health management according to an example embodiment is shown.

[0035] Figure 4 A block diagram of a computing device according to an example embodiment of the invention is shown. DETAILED DESCRIPTION

[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the present invention will be comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar parts, and thus their repeated description will be omitted.

[0037] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present invention. However, those skilled in the art will appreciate that the technical solution of the present invention can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known methods, devices, implementations or operations are not shown or described in detail to avoid blurring various aspects of the present invention.

[0038] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0039] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0040] It should be understood that although the terms first, second, third, etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another component. Therefore, the first component discussed below can be referred to as the second component without departing from the teachings of the present inventive concept. As used herein, the term "and / or" includes any one of the associated listed items and all combinations of one or more.

[0041] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0042] Those skilled in the art will appreciate that the drawings are merely schematic diagrams of example embodiments, and the modules or processes in the drawings are not necessarily necessary for implementing the present invention, and therefore cannot be used to limit the protection scope of the present invention.

[0043] Traditional biological asset supervision methods mainly rely on manual observation and simple mechanical equipment. Taking cattle as an example, this method requires long-term observation by ranch staff and manual supervision and tracking of cattle, which is not only labor-intensive but also inefficient. In addition, it is difficult for traditional methods to monitor the behavior and health status of each cow around the clock and in all directions, which may lead to some problems, such as diseases and abnormal behaviors cannot be discovered and handled in time. Traditional supervision often relies on subjective judgment, limited data collection and analysis capabilities, and lack of accurate data support, which limits the ability to conduct in-depth analysis and optimization of cattle health and production efficiency.

[0044] Therefore, to solve the above shortcomings of traditional biological asset supervision methods, a method of biological asset supervision based on knowledge graph is needed.

[0045] Before describing the embodiments of the present invention, some terms or concepts involved in the embodiments of the present invention are explained.

[0046] Transformer model, in computer terminology, is a deep learning model. The Transformer model is completely based on the attention mechanism, abandoning the previously common recurrent network (RNN) and convolutional network (CNN) structures, and focusing on processing sequence data.

[0047] BERT (Bidirectional Encoder Representations from Transformers) is a model based on the Transformer architecture. The innovation of BERT is that it uses a bidirectional Transformer to pre-train contextual deep language representations, which means that the representation of each word takes into account the information of all the words on its left and right.

[0048] TransD is a knowledge graph embedding method. TransD is designed to solve the high complexity problem of TransR / CTransR models when dealing with entity and relationship embedding. It improves the representation of entities and relationships through a dynamic mapping matrix instead of a fixed mapping matrix.

[0049] Token refers to the basic unit in the text data processing process. In the context of natural language processing (NLP) and computer science, when processing text, the original string data is divided into meaningful elements, which are tokens.

[0050] Knowledge graph technology is a structured semantic knowledge base that is designed to store relationships between entities (such as people, places, and things) and represent these entities and relationships through edges and nodes. The core of knowledge graph is to represent and understand complex data relationships through graphical structures. It is often used to improve the quality of search engine results and enhance the capabilities of expert systems. Knowledge graph has flexible scalability and can continuously add new entities and relationships as needed to continuously expand the knowledge base.

[0051] Exemplary embodiments of the present invention are described below with reference to the accompanying drawings.

[0052] Figure 1 A schematic diagram of an architecture for building a knowledge graph according to an example embodiment is shown.

[0053] According to some embodiments, the knowledge graph construction is to extract knowledge through empirical text data describing biological states, construct a knowledge graph and use knowledge graph embedding technology to capture semantic information for efficient calculation and reasoning.

[0054] See also Figure 1 , the construction of the biological behavior knowledge graph includes text processing, extracting word sequences from the text data to construct word sequences. Texts related to biological behavior often contain a large number of professional terms, abbreviations and complex concepts. Therefore, special attention should be paid to the correct processing of these professional vocabulary in the preprocessing stage. Sometimes customized dictionaries or word segmentation tools are required to improve the accuracy of word segmentation, divide the text into word sequences, assign a unique identifier to each word, and consider the word sequence characteristics unique to the biological field. Each word sequence is embedded in a vector, which includes tag embedding, position embedding and segment embedding. These embeddings are added to form the embedding vector of the word, and the embedding vector (which can be denoted as E) is input into the encoder for feature extraction.

[0055] According to some embodiments, in the construction of the biological behavior knowledge graph, the BERT model can deeply understand the biological behavior description, experimental results, gene function, etc. in the text, and extract feature vectors closely related to the biological behavior. The encoder includes a Transformer encoder based on the Bert model. The Bert-Base model used has a 12-layer transformer encoder, and the encoder includes a multi-head attention mechanism and a feedforward neural network. The embedded vector is input into the Transformer encoder for feature extraction, and a hidden layer feature vector with semantic features can be extracted.

[0056] The feature vector extracted by the encoder is passed to a linear classifier for entity recognition, and the last layer output of the BERT model (which can be denoted as H) is passed to a linear classifier, which predicts an entity category for each tag. The entity category of each tag is predicted by a linear transformation and a Softmax function, and the entity recognition can be achieved by using a cross entropy loss function.

[0057] According to an example embodiment, relationship extraction is performed on the identified entities, and the relationship extraction can also apply the BERT model. For each pair of identified entities, their vector representations are used to predict the relationship between them. In order to predict the relationship between two entities, the model combines the vector representations of the two entities. The combined feature vector includes the connection, difference, and element product of the entity vectors, aiming to capture the complex semantic relationship between entities. Parameters are optimized by minimizing the cross entropy loss function. An entity consists of multiple tokens, which record the start and end positions of the entity.

[0058] After the two steps of entity recognition and relationship extraction, the entities output by the module predict their relationships based on each pair of entities, and the entities and their relationships are constructed into triples, providing basic units for the construction of the knowledge graph. For each identified entity pair, its predicted relationship type together with the entity itself constitutes a triple (head entity, relationship, tail entity). All such triple sets can be used to construct a knowledge graph, in which entities are nodes and relationships are edges connecting these nodes, to construct a preliminary knowledge graph. The knowledge graph can represent facts and semantic relationships in the text in a structured way, providing a basis for subsequent reasoning, query and analysis.

[0059] According to an example embodiment, knowledge graph embedding is implemented based on the TransD model. The triples are used to represent the preliminary knowledge graph generated by the above process, which contains a set of entities E and relations R. A d-dimensional embedding vector is learned for each entity. For each relation R, a k-dimensional embedding vector is learned. In addition, each entity E has a projection vector and each relation R has a projection vector. Knowledge graph embedding projects the entity embedding vector into a relationship-related space through the TransD model.

[0060] This example embodiment can extract structured knowledge from massive bio-behavior text data, build bio-behavior and other knowledge graphs, and provide strong support for bio-behavior research, disease and health analysis, bio-regulation and other fields. The BERT-based relationship extraction method fully utilizes the powerful ability of the pre-trained model in understanding language context, effectively promotes the conversion from unstructured text to structured knowledge, and is one of the effective tools for knowledge graph construction and maintenance.

[0061] Figure 2 A flow chart of a method for biological asset management based on a knowledge graph according to an example embodiment is shown.

[0062] In S101, a knowledge graph is pre-constructed from text data describing empirical knowledge of biological assets, and the knowledge graph is completed by knowledge graph embedding.

[0063] According to an example embodiment, a word sequence is constructed from text data describing empirical knowledge of biological assets, the text data including veterinary reports, agricultural research papers, and / or daily ranch records. Each word forms a corresponding word embedding vector, which is input into an encoder for feature extraction, and the feature vector extracted by the encoder is passed to a linear classifier for entity recognition.

[0064] Relationship extraction is performed on the identified entities to obtain a knowledge graph including entities and relationships. For example, the entities and the relationships are constructed into triples as basic units for constructing the knowledge graph.

[0065] Then, the implicit relationship between entities in the knowledge graph is discovered through the knowledge graph embedding method, and the knowledge graph is completed based on the implicit relationship.

[0066] In S103, images of the biological assets are collected.

[0067] According to an example embodiment, video image data of the daily behavior of the biological asset is obtained. The methods for collecting video image data include but are not limited to installing fixed or mobile cameras, using drones for aerial photography, and setting automatic triggering cameras, such as infrared sensing cameras, to automatically take photos or videos when the biological asset enters the sensing range. In combination with the use of image and video analysis software, the behavior of specific organisms can be automatically identified and tracked, such as motion trajectory analysis, abnormal behavior detection, etc.

[0068] Video image data can be used to monitor animal health, behavioral patterns, growth cycles or reproductive habits to obtain key information.

[0069] In S105, basic behavior recognition is performed on the biological asset according to the collected image.

[0070] According to an example embodiment, the basic behavior recognition of the biological asset by the collected image includes using a convolutional neural network to recognize the behavior of the biological asset in the video image data. The collected video images are preprocessed to adapt to the requirements of the model input and improve the generalization ability of the model. In biological behavior recognition, CNN learns to capture the specific movements, postures or interaction patterns of animals. The trained CNN model can map the extracted features to specific behavior categories, and the supervised learning process involved is trained on a large number of labeled behavior instances.

[0071] For video data, in addition to analyzing static frames, it is also necessary to consider changes in behavior over time, and adjust the weights of the CNN through the back-propagation algorithm to minimize the difference between predicted behavior and actual behavior. Once the model training is completed, it can be applied to newly collected video data to automatically identify and classify the basic behavior of biological assets.

[0072] In S107 , a query vector is constructed using the identified behavior tags.

[0073] According to an example embodiment, constructing a query vector using the identified behavior tags is to convert the basic behavior information identified in step S105 into a form that can be calculated and compared, so as to facilitate query and reasoning of the knowledge graph in subsequent steps.

[0074] According to some embodiments, the behavior label may be a manually annotated behavior category, such as "eating", "running", "resting", etc. The identified behavior label is converted into a numerical vector representation. If a biological asset exhibits multiple behaviors within a certain period of time, the vector representations of these behaviors need to be aggregated to form a composite behavior vector, and the aggregated behavior vector is used as a query vector. The query vector contains comprehensive information about the behavior pattern of the biological asset, which can be used to find similar behavior patterns or related entities and relationships in the knowledge graph.

[0075] By constructing a query vector, the behavioral information of biological assets can be converted into a compact mathematical representation, which not only facilitates storage and transmission, but also makes subsequent reasoning and query operations feasible.

[0076] In S109, the query vector is used to perform reasoning operations in the knowledge graph, and the reasoning results are converted into a recognizable form and output as supervision results.

[0077] According to the example embodiment, the constructed query vector is used to search for the entity vector or relationship vector that is most similar to it in the knowledge graph. By comparing the distance or angle between the query vector and the entity vector in the knowledge graph, cases similar to the behavior pattern of biological assets can be found, or the next action that biological assets may take can be predicted. Based on the results of the similarity search, reasoning operations are performed to provide decision support for the supervision and protection of biological assets.

[0078] Converting the vector-form results obtained from the reasoning operation into understandable regulatory information or decision recommendations, and presenting the reasoning results to regulators or decision makers in recognizable forms such as charts, reports or alerts, will help to continuously improve the accuracy and practicality of the biological asset regulatory system.

[0079] The biological behavior knowledge graph construction method proposed in this example embodiment supervises biological assets based on knowledge graph and machine learning. By performing reasoning operations in the knowledge graph, it can not only automatically analyze and understand the behavior patterns of biological assets, but also predict potential risks and opportunities, and provide scientific basis and intelligent support for the protection, management and utilization of biological resources. It reflects the cutting-edge trend of the application of big data and AI technology in the field of biological sciences.

[0080] It not only improves the understanding and utilization of empirical knowledge of biological assets, but also provides powerful data analysis and decision support tools for fields such as bioinformatics, ecology, and agricultural science. By constructing query vectors, the behavioral information of biological assets can be converted into a compact mathematical representation, which not only facilitates storage and transmission, but also makes subsequent reasoning and query operations feasible.

[0081] Figure 3A flow chart of a method for objective reasoning for biological asset health management according to an example embodiment is shown.

[0082] In S201, a knowledge graph is constructed.

[0083] As mentioned above, we collect text data related to biological assets from multiple sources for data collection and preprocessing, apply natural language processing technology for entity recognition and relationship extraction, organize entities and relationships into a structured graph form, and form a knowledge graph, in which entities are nodes and relationships are edges, expressing the complex relationship network of biological assets. We use knowledge graph embedding technology to map entities and relationships into a low-dimensional vector space to supplement and improve the knowledge graph for subsequent calculations and analysis.

[0084] In S203, images are collected.

[0085] According to some embodiments, image acquisition is the starting point of the supervision process, which involves using various devices to capture images or video data of biological asset areas. High-resolution images can provide real-time status of biological assets, including important information such as growth, distribution, and health status. For example, for monitoring cattle, cameras are installed in animal feeding places to shoot the production process of target objects 24 hours a day, and frames are extracted according to the hardware processing capacity as the basis for subsequent data processing.

[0086] In S205, basic behavior is identified.

[0087] According to some embodiments, basic behavior recognition involves processing and analyzing collected image data to identify and classify basic behavior patterns or states of organisms. Computer vision technology and deep learning algorithms, such as convolutional neural networks, are applied to automatically identify specific objects or behavior patterns in images, and ResNet convolutional neural networks are used to identify the behavior of biological assets in videos, such as identifying the behavioral targets of cattle, including lying down, walking, jumping, ruminating, drinking water, licking fur, licking calves, mating, etc.

[0088] In S207, the supervision target result is inferred.

[0089] According to the example embodiment, the biological behavior knowledge graph constructed above is used for biological asset supervision. The knowledge graph is a structured data organization form that can represent various entities and the complex relationships between them. By integrating the results of image analysis with the knowledge graph for reasoning, the depth and breadth of reasoning can be increased, and the accuracy of recognition and reasoning can be enhanced by using background knowledge such as known ecological laws and species habits.

[0090] Combining the output of image analysis with the information in the knowledge graph, a comprehensive analysis is performed through the reasoning engine and machine learning model to infer the results of regulatory objectives or make recommendations. The reasoning process aims to extract valuable information from massive data to support scientific decision-making and rapid response. The final reasoning regulatory objective results are mapped to entities and relationships in the generated knowledge graph based on the input behavior labels to construct corresponding queries and execute outputs.

[0091] The method of biological asset supervision based on the knowledge graph in this example embodiment integrates the biological asset supervision target reasoning technology with modern information technology and deep knowledge of biology, providing strong technical support for ecological protection and resource management, and helping to achieve more accurate and efficient biodiversity protection and sustainable utilization.

[0092] The knowledge graph includes a knowledge graph of the correspondence between cattle behaviors and diseases. In order to better illustrate the purpose and advantages of the present invention, the invention content is further explained below with reference to examples. Taking the target object cattle as an example, a knowledge graph of cattle diseases is constructed, covering the whole process from raw data to knowledge graph embedding.

[0093] According to an example embodiment, the data for text processing is derived from text data of veterinary reports, agricultural research papers, and daily records of ranches. In the preprocessing stage, natural language processing technology is used to clean the text, including removing noise data, standardizing medical terms, sentence segmentation, and word segmentation, etc., and then constructing it into a word sequence. The word sequence is embedded into a vector and input into the Transformer encoder.

[0094] According to some embodiments, entity recognition uses a machine learning model BERT to identify key entities in the text, such as disease names, symptoms, treatment methods, cattle breeds and ages. Entity classification is performed to classify the identified entities into different categories, such as disease entities, treatment entities, anatomical part entities, etc.

[0095] According to some embodiments, the relationship extraction uses a deep learning method, such as a neural network Bert model based on an attention mechanism, to extract the relationship between entities from the text. The relationship extraction includes a relationship definition, which requires the identification of the relationship between entities, such as "cause" (disease causes symptoms), "treatment" (treatment method treats the disease), and "affect" (disease affects the site).

[0096] According to some embodiments, the knowledge graph is constructed by using nodes to represent entities and directed edges to represent the relationship between entities. Each node and edge can be accompanied by attributes, such as the definition of the entity, source information, source of evidence for the edge, etc. Storage and indexing uses the graph database Neo4j to store the knowledge graph for efficient query and update.

[0097] According to some embodiments, knowledge graph embedding uses graph embedding technology TransD to convert nodes and edges in the knowledge graph into low-dimensional vector representations. The embedded vectors are used for downstream tasks such as disease prediction, similar disease recommendation, symptom analysis, etc.

[0098] Taking the target object cattle as an example, the process of using image recognition and knowledge graph reasoning to judge the disease condition of cattle is as follows.

[0099] According to some embodiments, a knowledge graph of cattle is constructed, and then high-resolution cameras are installed to collect images of cattle, and a surveillance camera system is used to collect images of cattle's daily activities in the pasture to ensure that various behaviors of cattle can be clearly captured. The basic behavior recognition of cattle is performed by extracting frames from the collected videos, using image processing techniques such as denoising and contrast enhancement to improve the accuracy of subsequent processing, and applying the deep learning model ResNet to analyze video frames to identify specific behaviors of cattle, such as shaking heads, frequent lying down, abnormal running, etc.

[0100] According to some embodiments, a knowledge graph is input, and the input knowledge graph includes a knowledge graph of the correspondence between cattle behaviors and diseases, wherein the nodes in the graph include various behaviors and diseases, and the edges represent the relationship between behaviors and diseases, such as "shaking head" behavior may be related to "ear infection". Knowledge graph reasoning uses graph query and logical reasoning algorithms to query the knowledge graph based on the identified behavior patterns to infer possible health problems of cattle. If a cattle exhibits multiple behaviors associated with a certain disease, the system will increase the likelihood assessment of the disease.

[0101] This example embodiment implements the supervision of biological assets of beef cattle based on the knowledge graph, and realizes the supervision of biological assets of living beef cattle. It provides efficient and reliable digital management for the supervision of living assets of beef cattle, which is conducive to improving the extensive management model of the beef cattle breeding industry.

[0102] By constructing a universal regulatory framework, the present invention can quickly expand biological asset regulatory capabilities through knowledge graph construction, and improve the accuracy of biological asset regulation by utilizing existing knowledge and experience data.

[0103] Figure 4 A block diagram of a computing device according to an example embodiment of the invention is shown.

[0104] like Figure 4 As shown, computing device 30 includes processor 12 and memory 14. Computing device 30 may also include bus 22, network interface 16, and I / O interface 18. Processor 12, memory 14, network interface 16, and I / O interface 18 may communicate with each other via bus 22.

[0105] The processor 12 may include one or more general-purpose CPUs (Central Processing Units, processors), microprocessors, or application-specific integrated circuits, etc., for executing relevant program instructions. According to some embodiments, the computing device 30 may also include a high-performance graphics card (GPU) 20 for accelerating the processor 12.

[0106] The memory 14 may include a machine system readable medium in the form of a volatile memory, such as a random access memory (RAM), a read-only memory (ROM) and / or a cache memory. The memory 14 is used to store one or more programs including instructions and data. The processor 12 can read the instructions stored in the memory 14 to execute the above-mentioned method according to the embodiment of the present invention.

[0107] The computing device 30 may also communicate with one or more networks via the network interface 16. The network interface 16 may be a wireless network interface.

[0108] The bus 22 may include an address bus, a data bus, a control bus, etc. The bus 22 provides a path for exchanging information between components.

[0109] It should be noted that, in the specific implementation process, the computing device 30 may also include other components necessary for normal operation. In addition, those skilled in the art may understand that the above device may only include components necessary for implementing the embodiments of this specification, and need not include all components shown in the figure.

[0110] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor. The computer-readable storage medium may include, but is not limited to, any type of disk, including a floppy disk, an optical disk, a DVD, a CD-ROM, a microdrive, and a magneto-optical disk, a ROM, a RAM, an EPROM, an EEPROM, a DRAM, a VRAM, a flash memory device, a magnetic card or an optical card, a nanosystem (including a molecular memory IC), a network storage device, a cloud storage device, or any type of medium or device suitable for storing instructions and / or data.

[0111] An embodiment of the present invention also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all of the steps of any method recorded in the above method embodiments.

[0112] Those skilled in the art can clearly understand that the technical solution of the present invention can be implemented with the help of software and / or hardware. "Unit" and "module" in this specification refer to software and / or hardware that can independently complete or cooperate with other components to complete specific functions, where the hardware can be, for example, a field programmable gate array, an integrated circuit, etc.

[0113] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0114] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0115] In the several embodiments provided by the present invention, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0116] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0118] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention.

[0119] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0120] The exemplary embodiments of the present invention are specifically shown and described above. It should be understood that the present invention is not limited to the detailed structures, configurations or implementations described herein; on the contrary, the present invention is intended to cover various modifications and equivalent configurations included in the spirit and scope of the appended clauses.

Claims

1. A method for biological asset supervision based on knowledge graph, characterized in that: include: Pre-constructing a knowledge graph from text data describing empirical knowledge of biological assets, and completing the knowledge graph through knowledge graph embedding; Capturing images of the biological assets; Performing basic behavior recognition on the biological asset based on the collected images; Constructing a query vector using the identified behavior labels; The query vector is used to perform reasoning operations in the knowledge graph, and the reasoning results are converted into a recognizable form and output as supervision results.

2. The method according to claim 1, characterized in that: The pre-constructing of a knowledge graph from text data describing the empirical knowledge of biological assets and completing the knowledge graph by embedding the knowledge graph includes: The text data describing the experience knowledge of biological assets is constructed into word sequences, each word forms a corresponding word embedding vector, and the word embedding vector is input into the encoder for feature extraction; Passing the feature vector extracted by the encoder to a linear classifier for entity recognition; Extract relationships from identified entities to obtain a knowledge graph that includes entities and relationships; The knowledge graph embedding method is used to discover the implicit relationships between entities in the knowledge graph, and the knowledge graph is completed based on the implicit relationships.

3. The method according to claim 2, characterized in that The encoder includes: a Transformer encoder based on the Bert model.

4. The method according to claim 2, characterized in that: The image collection of the biological assets includes: Obtain video image data of the daily behavior activities of the biological asset.

5. The method according to claim 4, characterized in that The basic behavior recognition of the biological asset based on the collected images includes: A convolutional neural network is used to identify the behavior of the biological asset in the video image data.

6. The method according to claim 2, characterized in that Relationship extraction is performed on the identified entities to obtain a knowledge graph consisting of entities and relationships, including: The entities and the relationships are constructed into triples as basic units for constructing the knowledge graph.

7. The method according to claim 1, characterized in that The text data includes: veterinary reports, agricultural research papers, and / or farm daily records.

8. The method according to claim 1, characterized in that The knowledge graph includes a knowledge graph of the correspondence between cattle behaviors and diseases.

9. A computer program product, characterized in that The method comprises a computer program, which implements the method according to any one of claims 1 to 8 when being executed by a processor.

10. A computing device, characterized in that include: processor; as well as A memory storing a computer program, which, when executed by the processor, enables the processor to perform the method according to claims 1 to 9.