Artificial Intelligence-Based Big Data Mining Method and System for Aquaculture Feedback

Abstract

This application relates to the field of artificial intelligence technology, and involves a method and system for mining big data on breeding feedback based on artificial intelligence. This application accurately identifies target breeding feedback texts and their derivative sequences from a large amount of big data on breeding feedback, and constructs a global breeding feedback text sequence, effectively integrating the key information in the breeding process. The introduction of the target breeding feedback mining network, combined with the set breeding feedback semantic space sequence, realizes the in-depth mining and enhancement of breeding feedback semantics; the use of semantic space feature collaboration technology to generate a target breeding feedback semantic collaboration space not only strengthens the internal connection between texts, but also improves the sensitivity and accuracy of abnormal event recognition; finally, the correlation degree between the target breeding feedback text and the abnormal breeding trigger event is calculated through the target pairing neural network, providing timely and accurate abnormal early warning and cause analysis for breeders, which helps to quickly respond and take effective measures.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology. Specifically, it relates to a method and system for mining big data on aquaculture feedback based on artificial intelligence. Background Art

[0002] With the continuous expansion of the scale of modern aquaculture and the improvement of the intelligent level, the data accumulation in the aquaculture process shows an explosive growth. These data, especially the aquaculture feedback data, contain rich aquaculture status information, which is of great significance for timely discovering aquaculture problems, preventing disease outbreaks, and optimizing aquaculture management strategies. However, traditional aquaculture feedback data analysis methods often rely on manual experience and simple statistical means, making it difficult to efficiently extract valuable information from the massive and complex data, and being easily affected by subjective factors, resulting in insufficient accuracy and timeliness of the analysis results.

[0003] In the related art, directly applying existing artificial intelligence technologies to mining big data on aquaculture feedback still faces many challenges. On the one hand, aquaculture feedback texts usually have the characteristics of being unstructured, rich in semantics, and closely contextually related, requiring effective text processing technologies to extract and represent their deep semantic information; on the other hand, abnormal events in the aquaculture process often involve the interaction of multiple factors, and a single data source or analysis dimension is difficult to comprehensively reveal their internal mechanisms and influence scopes. Summary of the Invention

[0004] In view of the problems mentioned above, in combination with the first aspect of this application, embodiments of this application provide a method for mining big data on aquaculture feedback based on artificial intelligence. The method includes:

[0005] Obtaining a collaborative aquaculture feedback text sequence of a target aquaculture feedback text and a target abnormal aquaculture trigger event from big data on aquaculture feedback;

[0006] Determining a derivative aquaculture feedback text sequence of the target aquaculture feedback text from the collaborative aquaculture feedback text sequence of the target abnormal aquaculture trigger event;

[0007] Generating a global aquaculture feedback text sequence based on the derivative aquaculture feedback text sequence and the target aquaculture feedback text, and loading the global aquaculture feedback text sequence into a target aquaculture feedback mining network; the target aquaculture feedback mining network includes a set aquaculture feedback semantic space sequence, and the target aquaculture feedback mining network is generated by learning knowledge based on the collaborative aquaculture feedback text sequence of an example abnormal aquaculture trigger event. The set aquaculture feedback semantic space in the set aquaculture feedback semantic space sequence is the target aquaculture feedback semantic space of the aquaculture feedback text, and the target aquaculture feedback semantic space of the aquaculture feedback text is generated by strengthening the semantic association link of the basic aquaculture feedback semantic space of the aquaculture feedback text;

[0008] Using the target aquaculture feedback mining network, in the set aquaculture feedback semantic space sequence, retrieve the set aquaculture feedback semantic space corresponding to each aquaculture feedback text in the global aquaculture feedback text sequence, and perform semantic space feature collaboration on the retrieved set aquaculture feedback semantic spaces to generate the target aquaculture feedback semantic collaboration space of the global aquaculture feedback text sequence;

[0009] Obtain the aquaculture feedback semantic space of the target aquaculture feedback text and the abnormal aquaculture feature space of the target abnormal aquaculture trigger event, and load the target aquaculture feedback semantic collaboration space, the aquaculture feedback semantic space of the target aquaculture feedback text, and the abnormal aquaculture feature space of the target abnormal aquaculture trigger event into a target pairing neural network to generate the correlation degree between the target aquaculture feedback text and the target abnormal aquaculture trigger event.

[0010] On the other hand, an embodiment of the present application further provides an aquaculture feedback mining system, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present application significantly improves the processing efficiency and analysis accuracy of aquaculture feedback data. By accurately identifying the target aquaculture feedback text and its derivative sequences from a large amount of aquaculture feedback big data, and constructing a global aquaculture feedback text sequence, the key information in the aquaculture process is effectively integrated, providing a comprehensive and in-depth data basis for subsequent analysis. The introduction of the target aquaculture feedback mining network, combined with the set aquaculture feedback semantic space sequence, realizes the in-depth mining and enhancement of aquaculture feedback semantics, significantly enhancing the model's understanding ability of complex aquaculture situations. Further, using the semantic space feature collaboration technology to generate the target aquaculture feedback semantic collaboration space not only strengthens the internal connection between texts, but also improves the sensitivity and accuracy of abnormal event recognition. Finally, by calculating the correlation degree between the target aquaculture feedback text and the abnormal aquaculture trigger event through the target pairing neural network, timely and accurate abnormal early warning and cause analysis are provided for aquaculture personnel, which helps to quickly respond and take effective measures, reduce aquaculture losses, and improve aquaculture benefits. Thereby, the aquaculture management level is improved and the aquaculture safety is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic flowchart of the execution process of the aquaculture feedback big data mining method based on artificial intelligence provided by an embodiment of the present application.

[0013] Figure 2It is a schematic diagram of the hardware architecture of the breeding feedback mining system provided by the embodiments of the present application. Detailed implementation manners

[0014] The present application will be specifically described below with reference to the accompanying drawings of the specification. Figure 1 It is a schematic flowchart of a method for mining big data of breeding feedback based on artificial intelligence provided by an embodiment of the present application. The method for mining big data of breeding feedback based on artificial intelligence will be introduced in detail below.

[0015] Step S110, obtain a collaborative breeding feedback text sequence of target breeding feedback texts and target abnormal breeding trigger events from the big data of breeding feedback.

[0016] Specifically, the big data of breeding feedback refers to a collection of breeding environment data (such as temperature, dissolved oxygen, pH value, etc.) and feedback text data of breeding personnel collected in real time from multiple farms, which is usually stored in a big data warehouse for subsequent data mining and analysis. For example, assume that a large-scale pig breeding area has 10 farms, and each farm generates hundreds of pieces of breeding environment data and dozens of pieces of breeding feedback text every day. These data are transmitted to the server in real time and stored in the big data warehouse. The server can query this warehouse to obtain all the breeding feedback data within a specific time period for further analysis.

[0017] The target breeding feedback text refers to the breeding feedback text that is considered directly related to a specific abnormal breeding trigger event and has important analysis value under the background of the event. These texts usually contain key information, such as symptom descriptions, abnormal phenomena, etc. For example, in the abnormal breeding trigger event of "large-scale death of pigs", the target breeding feedback texts may include descriptions such as "a large number of dead pigs lying on the ground are found" and "the activity of the pig group suddenly decreases". These texts directly point to the core phenomena of the abnormal event, so they are selected as target texts for in-depth analysis. The target abnormal breeding trigger event refers to an abnormal event that occurs during the breeding process and may cause breeding problems or losses, such as disease outbreaks, water source deterioration, feed poisoning, etc. These abnormal events are usually identified through abnormal changes in environmental data or specific descriptions in breeding feedback texts. For example, during the breeding process, the server detects that the dissolved oxygen concentration in a certain farm suddenly decreases, and the feedback text of the breeding personnel mentions that "pigs show symptoms of difficulty breathing". Based on this information and a pre-trained abnormal detection model, the server identifies "abnormal dissolved oxygen concentration" as the target abnormal breeding trigger event.

[0018] The collaborative farming feedback text sequence refers to the collection of all relevant farming feedback texts collected within a certain time period before and after the occurrence of a target abnormal farming trigger event. These texts are arranged in chronological order to reflect the development process and context of the event. For example, after the event of "mass death of live pigs" occurred, the server retrieved all farming feedback texts from one week before the event to three days after the event and arranged them in chronological order to form a collaborative farming feedback text sequence. This sequence contains a series of texts from "decrease in the activity level of the pig herd" to "discovery of a large number of dead pigs", reflecting the development track of the event.

[0019] That is, in this embodiment, a large-scale pig farming area has multiple farms, and each farm is equipped with an intelligent monitoring system for real-time collection of farming environment data (such as temperature, dissolved oxygen, pH value, etc.) and recording of the feedback texts of the farming personnel. These farming feedback texts cover various situations in the farming process, such as the growth status of live pigs, the occurrence of diseases, the effect of feed feeding, etc. The server is responsible for mining valuable information from the farming feedback big data.

[0020] Specifically, the server first collects recent farming feedback text data from the monitoring systems of each farm through API interfaces or database queries. These farming feedback text data are stored in a big data warehouse and indexed according to the timestamp and farm identifier. Then, the server uses a pre-trained anomaly detection model to analyze the collected environment data to identify target abnormal farming trigger events that may cause farming problems, such as sudden drop in temperature, abnormal dissolved oxygen concentration, etc. Each target abnormal farming trigger event is assigned a unique identifier, and the time point of occurrence and relevant environment parameters are recorded. For each identified target abnormal farming trigger event, the server retrieves all farming feedback texts within a certain time period before and after the occurrence of the target abnormal farming trigger event (such as one week before the event to three days after the event) by querying the big data warehouse. These farming feedback texts are arranged in chronological order to form the collaborative farming feedback text sequence of the target abnormal farming trigger event. In the collaborative farming feedback text sequence, the server can screen out the farming feedback texts directly related to the target abnormal farming trigger event as the target farming feedback texts according to business rules or keyword matching. For example, if the target abnormal farming trigger event is "mass death of live pigs", the target farming feedback texts may include descriptions such as "decrease in the activity level of the pig herd" and "discovery of dead pigs".

[0021] Step S120, determine the derivative farming feedback text sequence of the target farming feedback text from the collaborative farming feedback text sequence of the target abnormal farming trigger event.

[0022] Specifically, a derivative farming feedback text sequence refers to a text sequence in the collaborative farming feedback text sequence that is directly or indirectly related to the target farming feedback text and provides more information about the development process of the target abnormal farming trigger event. These texts are usually filtered from the collaborative sequence through semantic relevance calculation. For example, still taking the previous explanation as an example, in the collaborative farming feedback text sequence of "large-scale death of live pigs", the server calculates the semantic relevance between each text and the target farming feedback text (such as "dead pigs found") and filters out several texts with the highest relevance (such as "decreased activity of the pig herd", "abnormal water source detection") as the derivative farming feedback text sequence.

[0023] That is to say, after determining the target farming feedback text, the server needs to further analyze the collaborative farming feedback text sequence to find other texts that are directly or indirectly related to the target farming feedback text, and these texts may provide more information about the development process of the target abnormal farming trigger event.

[0024] Specifically, for each collaborative farming feedback text in the collaborative farming feedback text sequence, the server first extracts its target farming feedback semantic space. This process involves converting the collaborative farming feedback text into a high-dimensional vector representation that captures the key semantic information in the collaborative farming feedback text. The target farming feedback semantic space is the result of strengthening the semantic association link in the basic semantic space and can more accurately reflect the internal relationship between texts.

[0025] Next, the server calculates the semantic relevance between the target farming feedback text and other texts in the collaborative farming feedback text sequence, which is usually achieved by calculating the cosine similarity between text vectors or using more complex semantic matching algorithms. According to the descending order of semantic relevance, the server selects several texts with the highest relevance to the target farming feedback text from the collaborative farming feedback text sequence as the derivative farming feedback texts. Then, they are sorted according to their original chronological order in the collaborative farming feedback text sequence to generate the derivative farming feedback text sequence of the target farming feedback text.

[0026] Step S130: Generate a global farming feedback text sequence based on the derived farming feedback text sequence and the target farming feedback text, and load the global farming feedback text sequence into the target farming feedback mining network. The target farming feedback mining network includes a set of farming feedback semantic space sequences. The target farming feedback mining network is generated by learning knowledge from the collaborative farming feedback text sequences of sample abnormal farming trigger events. The set farming feedback semantic space in the set of farming feedback semantic space sequences is the target farming feedback semantic space of the farming feedback text. The target farming feedback semantic space of the farming feedback text is generated by strengthening the semantic association links of the basic farming feedback semantic space of the farming feedback text.

[0027] Specifically, the global farming feedback text sequence refers to the complete text sequence formed by merging the target farming feedback text with its derived farming feedback text sequence. This sequence provides comprehensive information about the target abnormal farming trigger event for subsequent advanced semantic analysis and anomaly detection. For example, still taking the previous explanation as an example, in the event of "mass deaths of live pigs", the server merges the target farming feedback text "Dead pigs found" with its derived farming feedback text sequence (such as "Decreased activity of the pig herd", "Abnormal water source detection") into a global farming feedback text sequence. This sequence contains the complete information from the emergence of abnormal phenomena to the final result.

[0028] The target farming feedback mining network refers to a neural network model specifically designed for target farming feedback text analysis. By training a large number of collaborative farming feedback text sequences of sample abnormal farming trigger events, it learns the complex associations and temporal features between texts and generates high-level feature representations that can reflect the overall farming feedback situation. For example, the server trains a target farming feedback mining network using a large amount of historical data. This network can receive the global farming feedback text sequence as input, extract and collaboratively process text features through a multi-layer neural network structure, and finally generate a high-level feature vector or matrix that can comprehensively reflect the farming feedback situation.

[0029] The set of farming feedback semantic space sequences refers to a series of farming feedback semantic spaces defined within the target farming feedback mining network. Each set farming feedback semantic space corresponds to a specific farming feedback scenario or semantic concept, used to represent the key semantic information in the text. In the target farming feedback mining network, the set of farming feedback semantic space sequences may include "Health status space of live pigs", "Water source status space", "Feed effect space", etc. Each set farming feedback semantic space has obtained a corresponding vector representation or matrix form through pre-training or network learning, used to capture the semantic information related to this set farming feedback semantic space in the text.

[0030] The basic farming feedback semantic space refers to the original farming feedback semantic space that has not been strengthened by semantic association links and may only contain the basic semantic information in the text. The target farming feedback semantic space, on the other hand, is the result of strengthening the semantic association links in the basic space and can more accurately reflect the internal connections and complex semantic relationships between texts. For example, for a text describing "pigs showing difficulty breathing", its basic farming feedback semantic space may only contain the vector representations of the two words "pigs" and "difficulty breathing". The target farming feedback semantic space after strengthening by semantic association links may also contain complex semantic information such as the causal relationship and chronological order relationship between these words.

[0031] That is, in this embodiment, after obtaining the target farming feedback text and its derivative text sequence, the server combines them into a global farming feedback text sequence for more in-depth analysis. Then, the server loads this global farming feedback text sequence into a neural network specifically designed for target farming feedback mining.

[0032] Specifically, the server places the target farming feedback text at the beginning of the global farming feedback text sequence, followed immediately by the derivative farming feedback text sequence. If the derivative text sequence is long, the first N texts can be intercepted as needed to maintain the rationality of the sequence. The constructed global farming feedback text sequence is loaded into the target farming feedback mining network, which is trained on the collaborative farming feedback text sequences of a large number of sample abnormal farming trigger events and can capture the complex associations and temporal features between farming feedback texts. The target farming feedback mining network internally contains a set of farming feedback semantic space sequences, and each set farming feedback semantic space corresponds to a specific farming feedback scenario.

[0033] In the target farming feedback mining network, each text is first mapped to its corresponding set farming feedback semantic space. Then, the target farming feedback mining network performs feature extraction and collaborative processing on these spaces through a multi-layer neural network structure to generate a high-level feature representation that can reflect the global farming feedback situation.

[0034] Step S140, using the target farming feedback mining network, in the set of farming feedback semantic space sequences, retrieve the set farming feedback semantic spaces corresponding to each farming feedback text in the global farming feedback text sequence, and perform semantic space feature collaboration on the retrieved set farming feedback semantic spaces to generate the target farming feedback semantic collaboration space of the global farming feedback text sequence.

[0035] In this embodiment, in the target aquaculture feedback mining network, the server deeply processes each aquaculture feedback text in the global aquaculture feedback text sequence, extracts its corresponding set aquaculture feedback semantic space, and generates a collaborative space that can comprehensively reflect the semantic content of the entire sequence through feature collaboration. For each text in the global aquaculture feedback text sequence, the server uses the retrieval unit in the target aquaculture feedback mining network to find the most matching semantic space in the set aquaculture feedback semantic space sequence. This process may involve various technologies such as vector similarity calculation and hash table lookup. The retrieved set aquaculture feedback semantic spaces are sent to the collaborative unit for processing. The collaborative unit performs feature fusion and collaboration on these set aquaculture feedback semantic spaces through advanced algorithms such as attention mechanism and graph neural network to capture the interactions and influences between the set aquaculture feedback semantic spaces. After feature collaboration processing, the server generates a target aquaculture feedback semantic collaborative space, which is a high-dimensional vector or matrix containing the comprehensive semantic information of all texts in the global aquaculture feedback text sequence and can be used for subsequent tasks such as correlation analysis and anomaly detection.

[0036] Step S150, obtain the aquaculture feedback semantic space of the target aquaculture feedback text and the abnormal aquaculture feature space of the target abnormal aquaculture trigger event, and load the target aquaculture feedback semantic collaborative space, the aquaculture feedback semantic space of the target aquaculture feedback text, and the abnormal aquaculture feature space of the target abnormal aquaculture trigger event into the target pairing neural network to generate the correlation degree between the target aquaculture feedback text and the target abnormal aquaculture trigger event.

[0037] Finally, the server needs to evaluate the correlation degree between the target aquaculture feedback text and the target abnormal aquaculture trigger event. For this purpose, the semantic space of the target aquaculture feedback text, the feature space of the target abnormal aquaculture trigger event, and the semantic collaborative space of the global aquaculture feedback text sequence are obtained and input into the target pairing neural network for calculation.

[0038] Specifically, the server first obtains the aquaculture feedback semantic space of the target aquaculture feedback text and the abnormal aquaculture feature space of the target abnormal aquaculture trigger event through the previous processing steps. The abnormal aquaculture feature space may contain key information such as abnormal values of environmental parameters and time points of abnormal occurrences.

[0039] Next, the server loads the aquaculture feedback semantic space of the target aquaculture feedback text, the abnormal aquaculture feature space of the target abnormal aquaculture trigger event, and the target aquaculture feedback semantic collaboration space into the target pairing neural network. This target pairing neural network is specifically designed to calculate the correlation between the text and the target abnormal aquaculture trigger event and may include components such as multiple hidden layers, attention mechanisms, and fusion layers. In the target pairing neural network, spatial information from different sources is fused together, and a correlation score is generated through complex non-linear transformations. This correlation score reflects the closeness or causal relationship between the target aquaculture feedback text and the target abnormal aquaculture trigger event. Finally, the server outputs the correlation score, which can be used in subsequent decision-making support, anomaly warning, or report generation scenarios. For example, if the correlation score is high, it may mean that the feedback text is of great value for understanding the cause or impact of the target abnormal aquaculture trigger event and requires special attention from aquaculture personnel.

[0040] Based on the above steps, the embodiments of the present application significantly improve the processing efficiency and analysis accuracy of aquaculture feedback data. By accurately identifying the target aquaculture feedback text and its derivative sequences from a large amount of aquaculture feedback big data and constructing a global aquaculture feedback text sequence, the key information in the aquaculture process is effectively integrated, providing a comprehensive and in-depth data basis for subsequent analysis. The introduction of the target aquaculture feedback mining network, combined with the set aquaculture feedback semantic space sequence, realizes the in-depth mining and enhancement of aquaculture feedback semantics, significantly enhancing the model's ability to understand complex aquaculture scenarios. Further, the use of semantic space feature collaboration technology to generate the target aquaculture feedback semantic collaboration space not only strengthens the internal connection between texts but also improves the sensitivity and accuracy of anomaly event recognition. Finally, by calculating the correlation between the target aquaculture feedback text and the abnormal aquaculture trigger event through the target pairing neural network, timely and accurate anomaly warnings and cause analysis are provided for aquaculture personnel, helping to quickly respond and take effective measures, reduce aquaculture losses, and improve aquaculture benefits. Thereby, the aquaculture management level is improved and aquaculture safety is ensured.

[0041] In a possible implementation manner, step S120 includes:

[0042] Step S121, for the collaborative aquaculture feedback text in the collaborative aquaculture feedback text sequence of the target aquaculture feedback text and the target abnormal aquaculture trigger event, extract the target aquaculture feedback semantic space of the aquaculture feedback text. The target aquaculture feedback semantic space is generated by strengthening the semantic association link of the basic aquaculture feedback semantic space of the aquaculture feedback text.

[0043] Exemplarily, assume that a large-scale pig farming area has multiple farms with intelligent monitoring. The server is responsible for collecting breeding environment data and feedback texts from these farms. Recently, the server identified a target abnormal breeding trigger event - "mass death of pigs", and hopes to deeply analyze the cause and process of this event.

[0044] Specifically, the server has retrieved all the breeding feedback texts within one week before and after the occurrence of the target abnormal breeding trigger event "mass death of pigs", constituting a collaborative breeding feedback text sequence. Now, the server needs to extract the target breeding feedback semantic space for these collaborative breeding feedback texts.

[0045] For example, the server first preprocesses each text in the collaborative breeding feedback text sequence, including removing stop words, normalizing punctuation, etc. Then, the server uses a pre-trained semantic model (such as BERT) to convert each text into a high-dimensional vector representation, and this vector representation is the basic breeding feedback semantic space. To more accurately reflect the internal relationship between texts, the server inputs the basic breeding feedback semantic space into the target semantic association link reinforcement network. This target semantic association link reinforcement network performs reinforcement processing on the semantic space through specific algorithms (such as graph attention network, Transformer, etc.) to generate the target breeding feedback semantic space. This process may involve identifying key entities in the text (such as breeding varieties, disease names), constructing a relationship graph between entities, and strengthening these relationships through technologies such as graph neural networks.

[0046] Step S122, according to the semantic association degree between the target breeding feedback text and the target breeding feedback semantic spaces of the respective collaborative breeding feedback texts, and in the descending order of the semantic association degree, determine multiple derivative breeding feedback texts of the target breeding feedback text from the collaborative breeding feedback text sequence of the target abnormal breeding trigger event.

[0047] The server has extracted the target breeding feedback semantic space for each collaborative breeding feedback text in the collaborative breeding feedback text sequence. Next, the server needs to calculate the semantic association degree between the target breeding feedback text (such as "found dead pigs") and other texts in the sequence to determine the derivative breeding feedback texts.

[0048] Specifically, the server calculates the cosine similarity between the target farming feedback text and the target farming feedback semantic space of each text in the collaborative farming feedback text sequence, or uses a more complex semantic matching algorithm (such as similarity calculation based on the attention mechanism) to obtain the semantic correlation degree. According to the descending order of the semantic correlation degree, the server selects several texts with the highest correlation degree with the target farming feedback text from the collaborative farming feedback text sequence as the derivative farming feedback texts. For example, texts such as "The activity of the live pig group has decreased" and "Abnormal water source detection" may be selected.

[0049] Step S123: Sort each derivative farming feedback text according to its collaborative time sequence position with the target abnormal farming trigger event to generate a derivative farming feedback text sequence of the target farming feedback text.

[0050] In this embodiment, the server has determined multiple derivative farming feedback texts of the target farming feedback text. Now, the server needs to sort these derivative farming feedback texts according to their original time order in the collaborative farming feedback text sequence to generate a derivative farming feedback text sequence.

[0051] Specifically, the server queries the timestamp of each derivative farming feedback text in the big data warehouse and sorts the derivative farming feedback texts according to the order of the timestamps. After the sorting is completed, the server combines the sorted derivative farming feedback texts into a sequence as the derivative farming feedback text sequence of the target farming feedback text. This derivative farming feedback text sequence reflects the possible changes and event development trajectories in the farming process from before the occurrence of the target abnormal farming trigger event to the appearance of the target farming feedback text.

[0052] Through the above steps, the server determines the derivative farming feedback text sequence of the target farming feedback text "Dead pigs found" from the collaborative farming feedback text sequence of the target abnormal farming trigger event "Large-scale death of live pigs". This derivative farming feedback text sequence may include texts such as "The activity of the live pig group has decreased" and "Abnormal water source detection", which are arranged in chronological order and provide important clues for understanding the development process of the target abnormal farming trigger event. These derivative farming feedback texts and their associated relationships will be further used for the construction of the global farming feedback text sequence and the generation of the target farming feedback semantic collaborative space, ultimately helping the server evaluate the correlation degree between the target farming feedback text and the target abnormal farming trigger event.

[0053] In a possible implementation manner, step S121 includes:

[0054] Step S1211: Generate a set of current farming feedback label attributes based on the farming feedback label attributes of the current farming feedback text, and load the set of current farming feedback label attributes into the target semantic association link strengthening network. The current farming feedback text is the target farming feedback text or the collaborative farming feedback text in the collaborative farming feedback text sequence of the target abnormal farming trigger event, and the target semantic association link strengthening network includes a sequence of basic farming feedback semantic spaces.

[0055] Exemplarily, the server in a large-scale pig farming area is responsible for processing and analyzing real-time data from multiple farms, including farming environment data and feedback text from farmers. The server has identified a target abnormal farming trigger event (such as "mass death of pigs") and retrieved the relevant collaborative farming feedback text sequence. Now, the server needs to extract the target farming feedback semantic space for these collaborative farming feedback texts (including the target farming feedback text).

[0056] Specifically, for each current farming feedback text in the collaborative farming feedback text sequence, the server first identifies and extracts its farming feedback label attributes, which may include farming variety, disease name, symptom description, feed type, etc.

[0057] For example, the server uses natural language processing techniques (such as named entity recognition, keyword extraction, etc.) to analyze the text, and identifies the key entities and descriptive words therein as farming feedback label attributes. Then, all the identified label attributes are organized into a set, that is, the set of current farming feedback label attributes. For example, for a text describing "decreased activity of pigs", its label attribute set may include {"pigs", "decreased activity"}.

[0058] Step S1212: Using the target semantic association link strengthening network, in the sequence of basic farming feedback semantic spaces, retrieve the basic farming feedback semantic space corresponding to the farming feedback label attributes of the current farming feedback text, and strengthen the semantic association link of the retrieved basic farming feedback semantic space to generate the target farming feedback semantic space of the current farming feedback text. <http: / / www.wipo.int / standards / XMLSchema / ST96 / ST96-EN.xml#

[0059] The server has constructed a target semantic association link strengthening network, which is specifically used to strengthen the semantic association link of farming feedback text to generate a more accurate target farming feedback semantic space. Now, the server needs to load the set of current farming feedback label attributes into this target semantic association link strengthening network.

[0060] Specifically, ensure that the target semantic association link reinforcement network has been trained and is in an available state. Inside the target semantic association link reinforcement network, there is a sequence of basic aquaculture feedback semantic spaces, which are predefined and used to represent different types of aquaculture feedback scenarios. Then, load the current aquaculture feedback label attribute set as input data into the target semantic association link reinforcement network. This usually involves converting the label attributes into a format understandable by the network (such as vector representation) and passing it to the input layer of the network.

[0061] In the target semantic association link reinforcement network, the server needs to use the set of label attributes of the current aquaculture feedback text to retrieve the corresponding basic aquaculture feedback semantic spaces and perform semantic association link reinforcement processing on these spaces. For example, the retrieval unit in the target semantic association link reinforcement network looks for the most matching semantic space in the sequence of basic aquaculture feedback semantic spaces according to the current aquaculture feedback label attribute set. This process may involve various techniques such as vector similarity calculation and hash table lookup. For the retrieved basic aquaculture feedback semantic spaces, the semantic association link reinforcement unit in the target semantic association link reinforcement network will perform reinforcement processing on them. This usually involves analyzing the association relationships between label attributes (such as causal relationships, parallel relationships, etc.) and using techniques such as graph neural networks and attention mechanisms to enhance the representation of these relationships in the semantic space. After the reinforcement processing, the target semantic association link reinforcement network generates the target aquaculture feedback semantic space of the current aquaculture feedback text. This target aquaculture feedback semantic space is a high-dimensional vector or matrix that can more accurately reflect the key semantic information in the text and the internal connection between label attributes.

[0062] Through the above steps, the server extracts the target aquaculture feedback semantic space for each current aquaculture feedback text in the collaborative aquaculture feedback text sequence. This process involves the identification and set construction of label attributes, data loading into the target semantic association link reinforcement network, and the retrieval and reinforcement processing of semantic spaces. The finally generated target aquaculture feedback semantic space provides important support for subsequent tasks such as semantic association degree calculation and anomaly detection.

[0063] In a possible implementation, the method further includes:

[0064] Step A110, obtain the first aquaculture feedback label attribute set and input the first aquaculture feedback label attribute set into the basic semantic association link reinforcement network. The first aquaculture feedback label attribute set is generated based on the collaborative aquaculture feedback text sequence of the first sample abnormal aquaculture trigger event. The basic semantic association link reinforcement network includes the sequence of basic aquaculture feedback semantic spaces.

[0065] In this embodiment, the server can select a representative first sample abnormal breeding trigger event (such as "mass death of pigs"). According to this first sample abnormal breeding trigger event, the server retrieves a sequence of collaborative breeding feedback texts within a period of time before and after the event. Next, using natural language processing techniques and a domain knowledge base, the server extracts key breeding feedback label attributes from the sequence of collaborative breeding feedback texts, such as breeding variety, symptom description, environmental parameter changes, etc., and constructs a first set of breeding feedback label attributes.

[0066] Thus, the server inputs the first set of breeding feedback label attributes into a pre-constructed basic semantic association link reinforcement network. The basic semantic association link reinforcement network internally contains a sequence of basic breeding feedback semantic spaces for representing different types of breeding feedback scenarios.

[0067] For example, first ensure that the basic semantic association link reinforcement network has been initialized and is in a trainable state. Then, load the first set of breeding feedback label attributes into the input layer of the network in a certain format (such as a vector sequence).

[0068] Step A120, using the basic semantic association link reinforcement network, in the sequence of basic breeding feedback semantic spaces, retrieve the basic breeding feedback semantic spaces corresponding to each breeding feedback label attribute before the first label node in the first set of breeding feedback label attributes, perform semantic association link reinforcement on the retrieved basic breeding feedback semantic spaces, generate an estimated breeding feedback semantic space corresponding to the first label node, and based on the estimated breeding feedback semantic space corresponding to the first label node, generate a first estimated confidence level for the breeding feedback label attribute on the first label node in the first set of breeding feedback label attributes. The first label node is determined from each breeding feedback label node in the first set of breeding feedback label attributes, and the first estimated confidence level is used to reflect the estimated derivative metric value between the breeding feedback text corresponding to the first label node and the forward breeding feedback text sequence corresponding to the first label node.

[0069] Step A130, generate a first training error based on the first estimated confidence levels respectively corresponding to the breeding feedback label attributes on each first label node in the first set of breeding feedback label attributes.

[0070] Step A140, update the weight and bias information of the basic semantic association link reinforcement network according to the first training error until the first training termination requirement is met, and generate the target semantic association link reinforcement network.

[0071] In this embodiment, in the basic semantic association link reinforcement network, the server traverses each tag node in the first breeding feedback tag attribute set. For each first tag node, the server retrieves the basic breeding feedback semantic spaces corresponding to all its previous tag nodes, and performs semantic association link reinforcement processing on these basic breeding feedback semantic spaces.

[0072] For the first tag node, the retrieval unit in the network searches for the corresponding basic breeding feedback semantic space among its previous tag nodes. Using the semantic association link reinforcement unit, the retrieved all basic breeding feedback semantic spaces are strengthened to capture the internal connections and temporal characteristics between tag attributes. After the strengthening process, an estimated breeding feedback semantic space corresponding to the first tag node is generated.

[0073] Based on the estimated breeding feedback semantic space corresponding to the first tag node, the server generates a first estimated confidence level corresponding to the breeding feedback tag attribute on this first tag node, and this first estimated confidence level reflects the estimated derivative metric value between this tag attribute and its forward breeding feedback text sequence. For example, the prediction unit in the network uses the estimated breeding feedback semantic space to calculate the first estimated confidence level through a specific algorithm (such as logistic regression, neural network output layer, etc.).

[0074] The server compares the first estimated confidence levels on all first tag nodes in the first breeding feedback tag attribute set with a certain metric standard (such as cross-entropy loss) between the actual tag attributes to generate a first training error. Then, the server uses this error to update the weight and bias information of the basic semantic association link reinforcement network.

[0075] For example, the first estimated confidence level can be compared with the actual tag attribute to calculate the first training error. Using optimization techniques such as the backpropagation algorithm, the weight and bias information of the network are updated according to the first training error, and this process may involve multiple iterations until the training error is reduced below a certain threshold or a preset number of training rounds is reached.

[0076] After sufficient training iterations, the performance of the basic semantic association link reinforcement network reaches a stable state. At this time, the server regards it as the target semantic association link reinforcement network and uses it for subsequent real-time data processing and analysis tasks.

[0077] During the training process, the performance of the network is regularly evaluated to ensure that its performance on the validation set meets expectations. Once the network meets the training termination requirements (such as the error is lower than the threshold, the training rounds are completed, etc.), the server saves it as the target semantic association link reinforcement network for subsequent use.

[0078] Through the above steps, the server trains the basic semantic association link reinforcement network and generates the target semantic association link reinforcement network. This process involves multiple links such as the selection of sample events, the retrieval of text sequences, the extraction of label attributes, the input and output processing of the network, the calculation of training errors, and the update of the network. The finally generated target network can more accurately capture the semantic associations and temporal features between aquaculture feedback texts, providing strong support for subsequent tasks such as aquaculture anomaly detection and cause analysis.

[0079] In a possible implementation manner, step A120 includes:

[0080] Step A121, perform semantic feature domain conversion on the estimated aquaculture feedback semantic space corresponding to the first label node to generate the basic semantic feature domain vector distribution corresponding to the first label node. The basic semantic feature domain vector distribution includes semantic feature domain vectors corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence.

[0081] In this embodiment, the server has generated the corresponding estimated aquaculture feedback semantic space for the first label node through the basic semantic association link reinforcement network. Now, the server needs to convert this estimated aquaculture feedback semantic space into the semantic feature domain for further analysis and processing.

[0082] Specifically, the server first parses the estimated aquaculture feedback semantic space and extracts the key semantic information and feature vectors contained therein. Then, using predefined mapping rules or models, the server maps these feature vectors into the semantic feature domain to generate the basic semantic feature domain vector distribution, and each vector in this basic semantic feature domain vector distribution represents the semantic feature of a certain reference aquaculture feedback text in the reference aquaculture feedback text sequence.

[0083] Step A122, perform semantic dependence reinforcement on the basic semantic feature domain vector distribution to generate the target semantic feature domain vector distribution corresponding to the first label node. The target semantic feature domain vector distribution includes the first estimated confidence levels corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence, and the reference aquaculture feedback text sequence includes the aquaculture feedback texts corresponding to each aquaculture feedback label attribute in the first aquaculture feedback label attribute set.

[0084] After obtaining the basic semantic feature domain vector distribution, the server needs to further strengthen the semantic dependence between these vectors to more accurately reflect the internal connections and temporal features between the reference aquaculture feedback texts.

[0085] Specifically, the server analyzes the potential dependency relationships between various vectors in the basic semantic feature domain vector distribution, such as causal relationships, parallel relationships, etc. Then, using advanced algorithms such as graph neural networks and attention mechanisms, the server performs reinforcement processing on these vectors to enhance the semantic dependencies between them. The processed vector distribution is the target semantic feature domain vector distribution.

[0086] Step A123, from the target semantic feature domain vector distribution, determine the first estimated confidence corresponding to the aquaculture feedback label attribute on the first label node in the first aquaculture feedback label attribute set.

[0087] After the semantic dependency reinforcement, the server obtains the target semantic feature domain vector distribution. Now, the server needs to determine the first estimated confidence corresponding to the aquaculture feedback label attribute on the first label node in the first aquaculture feedback label attribute set from this distribution.

[0088] For example, the server first finds the vector corresponding to the reference aquaculture feedback text that matches the aquaculture feedback label attribute on the first label node in the target semantic feature domain vector distribution. Since each vector in the target semantic feature domain vector distribution contains an estimated confidence (which can be directly generated by a certain output layer of the network or indirectly inferred through the relationship between vectors), the server directly extracts the corresponding first estimated confidence from this matching vector.

[0089] Finally, record the extracted first estimated confidence for subsequent training error calculation and network weight update.

[0090] Through the above steps, the server generates the first estimated confidence for the aquaculture feedback label attribute on the first label node. This process involves multiple links such as the conversion of the semantic feature domain, the reinforcement of semantic dependencies, and the extraction of confidence. The first estimated confidence reflects the accuracy of the network's estimation of the relationship between the label attribute and its forward aquaculture feedback text sequence, and is an important basis for evaluating the network performance and guiding network training. In subsequent iterative training, the server will continuously adjust the network weights and bias information according to these confidences to gradually improve the accuracy and generalization ability of the network.

[0091] In a possible implementation manner, the basic semantic association link reinforcement network includes a retrieval unit, a mapping unit, a semantic association link reinforcement unit, and a prediction unit. The retrieval unit is used to retrieve the basic aquaculture feedback semantic space, the mapping unit is used for semantic association link mapping, the semantic association link reinforcement unit is used for semantic association link reinforcement, and the prediction unit is used to generate the first estimated confidence.

[0092] Step A140 includes: updating the weights and bias information of the mapping unit, semantic association link reinforcement unit, and prediction unit in the basic semantic association link reinforcement network according to the first training error until the first training termination requirement is met, and generating the target semantic association link reinforcement network.

[0093] In this embodiment, in the intelligent monitoring system of large-scale pig breeding areas, the server is responsible for training and optimizing the basic semantic association link reinforcement network to construct a target network that can accurately understand and process the complex semantic associations between breeding feedback texts. This process involves the collaborative work of multiple network components, including the retrieval unit, mapping unit, semantic association link reinforcement unit, and prediction unit.

[0094] Specifically, the basic semantic association link reinforcement network consists of multiple specially designed units, and each unit undertakes a specific task.

[0095] Retrieval unit: responsible for retrieving the semantic space related to the currently processed label node from the basic breeding feedback semantic space sequence.

[0096] Mapping unit: maps the retrieved semantic space into a feature domain suitable for further processing.

[0097] Semantic association link reinforcement unit: enhances the semantic dependence between vectors in the feature domain to reflect the internal connection between texts.

[0098] Prediction unit: generates an estimated confidence level for the currently processed label node based on the reinforced feature domain vector distribution.

[0099] During the training process, the server calculates the first training error based on the difference between the estimated confidence level generated by the network and the actual label attribute. For example, the server compares the first estimated confidence level generated by the prediction unit with the actual label attribute obtained through a certain method (such as manual annotation, historical data verification, etc.). Using a predefined loss function (such as cross-entropy loss, mean square error, etc.), the server calculates the difference between the first estimated confidence level and the actual label attribute to obtain the first training error.

[0100] After obtaining the first training error, the server uses this first training error to guide the update of the network weights and bias information to optimize the network performance. The server adopts the backpropagation algorithm to propagate the first training error layer by layer backward from the prediction unit until the mapping unit and the retrieval unit. During the backpropagation process, the server calculates the adjustment amount of the weights and biases in each network unit according to the error gradient and applies these adjustment amounts to update the network weights and bias information.

[0101] Iterative optimization: This process is repeated multiple times. In each iteration, new training data is used to generate new estimated confidence levels and training errors, and based on these, the network weights and biases are further adjusted. As the iterations progress, the performance of the network gradually improves.

[0102] After a certain number of iterative training sessions, if the performance of the network reaches a preset standard or the training error is reduced to an acceptable range, the server will stop the training process and save the current network state as the target semantic association link reinforcement network. At the end of each iteration, the server evaluates the performance of the network (such as accuracy, recall, F1-score, etc.) and compares it with the preset training termination requirements. If the network performance meets the training termination requirements or the number of training rounds reaches the preset upper limit, the server will stop the training process. Finally, the server saves the currently optimized basic semantic association link reinforcement network as the target semantic association link reinforcement network for subsequent real-time data processing and analysis tasks.

[0103] Through the above steps, the server trains and optimizes the basic semantic association link reinforcement network, generating a target network that can accurately capture the complex semantic associations between aquaculture feedback texts. This process involves multiple aspects such as the collaborative work of network components, the calculation of training errors, the update of network weights and bias information, and the judgment of training termination conditions. The generation of the target network not only improves the accuracy and efficiency of the intelligent monitoring system but also provides more reliable and useful data analysis support for aquaculture personnel.

[0104] In a possible implementation, the method further includes:

[0105] Step B110, obtaining the text knowledge point distribution corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence.

[0106] Step B120, through a graph self-attention processing network, performing graph self-attention semantic mining on the text knowledge point distribution to generate the basic aquaculture feedback semantic space corresponding to each reference aquaculture feedback text.

[0107] Step B130, generating a sequence of the basic aquaculture feedback semantic spaces based on the basic aquaculture feedback semantic spaces corresponding to each reference aquaculture feedback text.

[0108] In this embodiment, the server retrieves a series of reference aquaculture feedback texts from the stored historical data, which cover aquaculture situations under different times and different aquaculture conditions. The server then performs natural language processing on each reference aquaculture feedback text to extract the knowledge point distribution in the reference aquaculture feedback text.

[0109] Specifically, the server retrieves a reference aquaculture feedback text sequence from the database according to preset query conditions (such as time range, aquaculture species, aquaculture problems, etc.). For each reference aquaculture feedback text in the reference aquaculture feedback text sequence, the server extracts key entities and descriptive words in the text as knowledge points using techniques such as named entity recognition and keyword extraction. These knowledge points can include aquaculture species, disease names, symptom descriptions, environmental parameters, etc. The server organizes the extracted knowledge points into a distribution form for subsequent semantic mining. This distribution can be a knowledge point set, a knowledge graph, or any other data structure suitable for representing text content.

[0110] Next, the server uses a pre-trained graph self-attention processing network to perform semantic mining on the knowledge point distribution of the reference aquaculture feedback text. This process aims to capture the internal connections and context information between knowledge points to generate a more accurate and rich semantic representation.

[0111] Specifically, ensure that the graph self-attention processing network has been trained and is in an available state. This network may include components such as multiple graph convolutional layers, self-attention layers, and fully connected layers. The server converts the knowledge point distribution of each reference aquaculture feedback text into an input format acceptable to the network (such as graph structure data, node feature vectors, etc.) and inputs it into the network. In the graph self-attention processing network, node features are propagated and aggregated through graph convolutional layers, and the self-attention layer is used to capture the complex dependencies between nodes. After multiple layers of processing, the network finally outputs the basic aquaculture feedback semantic space corresponding to each text. This space is a high-dimensional vector or matrix that contains the semantic information of the key knowledge points in the text and the association relationships between them.

[0112] Finally, the server organizes the basic aquaculture feedback semantic spaces corresponding to each reference aquaculture feedback text in the order in which they appear in the original text sequence to form an ordered basic aquaculture feedback semantic space sequence. This basic aquaculture feedback semantic space sequence provides important data support for subsequent advanced semantic analysis and anomaly detection tasks. The server sorts the generated basic aquaculture feedback semantic spaces according to the timestamps of the reference aquaculture feedback text sequence or other sorting criteria (such as aquaculture cycle, processing priority, etc.). The sorted basic aquaculture feedback semantic spaces are sequentially added to the sequence to form a complete basic aquaculture feedback semantic space sequence. This basic aquaculture feedback semantic space sequence can be saved in the database for subsequent use or directly passed to other processing modules for further analysis.

[0113] Through the above steps, the server extracts the text knowledge point distribution from the reference aquaculture feedback text and conducts in-depth semantic mining through the graph self-attention processing network. The finally generated basic aquaculture feedback semantic space sequence not only contains rich semantic information but also maintains the temporal relationship and internal connection between texts, providing a solid foundation for subsequent complex aquaculture data analysis, anomaly detection, and decision support.

[0114] In a possible implementation manner, the steps of obtaining the text knowledge point distribution corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence, and performing graph self-attention semantic mining on the text knowledge point distribution through a graph self-attention processing network to generate the basic aquaculture feedback semantic space corresponding to each reference aquaculture feedback text include:

[0115] Step C110, perform word segmentation on each reference aquaculture feedback text in the reference aquaculture feedback text sequence, segment the continuous reference aquaculture feedback text into a word segmentation sequence, and after performing part-of-speech tagging on the word segmentation sequence, use named entity recognition technology to extract the key entity sequence in the word segmentation sequence. The key entities in the key entity sequence include aquaculture varieties, disease names, and symptom descriptions.

[0116] In this embodiment, the server first performs word segmentation on each reference aquaculture feedback text in the reference aquaculture feedback text sequence, segmenting the continuous reference aquaculture feedback text into individual word segmentation sequences. Subsequently, the server performs part-of-speech tagging on these word segmentation sequences to distinguish words of different parts of speech such as nouns, verbs, and adjectives.

[0117] Specifically, the server uses a pre-trained word segmentation model (such as jieba word segmentation) to segment the text to obtain a word segmentation sequence. A part-of-speech tagging tool (such as HanLP) is used to perform part-of-speech tagging on each word in the word segmentation sequence, providing a basis for subsequent named entity recognition.

[0118] After obtaining the word segmentation sequence and part-of-speech tagging, the server uses named entity recognition technology to extract the key entity sequence from the word segmentation sequence. These key entities include entities such as aquaculture varieties, disease names, and symptom descriptions that are crucial for understanding the text content. For example, the server applies a named entity recognition model (such as BiLSTM-CRF) to scan the word segmentation sequence to identify the key entities in the text. The identified key entities are organized into a key entity sequence in the order they appear in the text.

[0119] Step C120: Extract the core vocabulary in the key entity sequence as knowledge points through a keyword extraction algorithm. Use the extracted knowledge points as nodes, establish edges according to the co-occurrence relationship of the nodes in the corresponding reference breeding feedback text, construct a knowledge point graph, and assign an initial vector representation to each node in the knowledge point graph.

[0120] In this embodiment, the server extracts the core vocabulary from the key entity sequence as knowledge points through a keyword extraction algorithm, and constructs a knowledge point graph with these knowledge points as nodes. The edges in the graph represent the co-occurrence relationship between nodes (knowledge points). For example, keyword extraction algorithms such as TF-IDF and TextRank can be used to extract the core vocabulary from the key entity sequence. Use the extracted knowledge points as nodes and establish edges according to their co-occurrence relationship in the reference breeding feedback text. The co-occurrence relationship can be determined by counting the number of times the nodes appear simultaneously in the text. Then, assign an initial vector representation to each node in the knowledge point graph. This vector can be random or obtained through a pre-trained word embedding model (such as Word2Vec, GloVe).

[0121] Step C130: Input the initial vector representation of each node into an embedding layer to obtain target feature embedding data. For each node in the knowledge point graph, calculate the attention score of this node to all other nodes. The attention score reflects the degree of dependence of the current node on other nodes. The calculation of the attention score is achieved through dot product, concatenation and then through a neural network, and the softmax function is used to normalize the attention score to obtain the attention weight of each node to other nodes.

[0122] Step C140: According to the attention weight, perform weighted aggregation on the features of the neighbor nodes of each node to obtain the updated feature representation of the current node. After linear transformation and non-linear activation of the aggregated updated feature representation, generate the target feature representation corresponding to each node. For each reference breeding feedback text, integrate the target feature representations of its corresponding nodes, and map the integrated target feature representation to the basic breeding feedback semantic space through a fully connected layer.

[0123] In this embodiment, the server uses a graph self-attention processing network to perform semantic mining on the knowledge point graph, updates the feature representation of the nodes by calculating the attention scores between the nodes and performing feature aggregation, and finally generates the basic breeding feedback semantic space corresponding to each reference breeding feedback text.

[0124] Specifically, the initial vector representation of each node can be input into an embedding layer (such as a multi-layer perceptron, a graph convolutional network, etc.) to obtain the target feature embedding data. For each node in the knowledge point graph, calculate the attention scores of this node with all other nodes, and this score reflects the degree of dependence of the current node on other nodes. The calculation of the attention scores can be achieved through dot product, concatenation followed by a neural network, and normalized using the softmax function to obtain the attention weights. According to the attention weights, perform weighted aggregation on the feature representations of the neighbor nodes of each node to obtain the updated feature representation of the current node, and this process takes into account the complex dependence relationships between nodes. Perform a linear transformation and non-linear activation (such as the ReLU activation function) on the aggregated updated feature representation to generate the target feature representation corresponding to each node. For each reference breeding feedback text, integrate the target feature representations of all its corresponding nodes (such as summation, averaging, etc.), and map the integrated feature representation through a fully connected layer to the basic breeding feedback semantic space, which is a high-dimensional vector or matrix containing the semantic information of the key knowledge points in the text and the association relationships between them.

[0125] Through the above steps, the server extracts the text knowledge point distribution from the reference breeding feedback text and conducts in-depth semantic mining through the graph self-attention processing network. The finally generated basic breeding feedback semantic space not only captures the key knowledge point information in the text, but also takes into account the complex dependence relationships and context information between the knowledge points. These basic breeding feedback semantic spaces provide strong data support for subsequent advanced semantic analysis, anomaly detection, and decision-making support.

[0126] Furthermore, in a practical example, assume that a large-scale pig breeding enterprise has deployed an intelligent monitoring system in multiple of its farms to monitor water source parameters (such as temperature, pH value, dissolved oxygen content) in real time and record the feedback texts of the breeding personnel. Recently, the enterprise has noticed that a certain pig disease (such as "bacterial septicemia") has occurred frequently in multiple farms, so it decides to use the server to conduct in-depth analysis of the historical breeding feedback texts to extract key knowledge points and construct the basic breeding feedback semantic space to provide support for subsequent disease early warning and prevention and control.

[0127] The server obtains a reference breeding feedback text: "Recently, it has been found that the activity of the black pigs in the pigsty has decreased, and some pigs have red spots on their body surfaces, suspected of being the initial symptoms of bacterial septicemia."

[0128] The server uses a pre-trained word segmentation model to segment the text and obtains the word segmentation sequence: "Recently found that the activity of the black pigs in the pigsty has decreased, and some pigs have red spots on their body surfaces, suspected of being the initial symptoms of bacterial septicemia."

[0129] Using a part-of-speech tagging tool (such as HanLP), the server performs part-of-speech tagging on the tokenized sequence, and the results can include nouns (such as "black pig", "red spot"), verbs (such as "decrease", "appear"), adjectives (such as "initial stage"), etc.

[0130] The server applies a named entity recognition model (such as BiLSTM-CRF) to scan the tokenized sequence and identifies key entities: "black pig" (breeding variety), "red spot" (symptom description), "bacterial septicemia" (disease name). These entities are organized into a key entity sequence in the order they appear in the text.

[0131] The server extracts core words from the key entity sequence as knowledge points: "black pig", "red spot", "bacterial septicemia". Then, a knowledge point graph is constructed with these knowledge points as nodes.

[0132] The server counts the co-occurrence times of these knowledge points in the reference breeding feedback text sequence. For example, "black pig" and "red spot" may appear simultaneously in multiple texts, indicating an association between them. Based on these co-occurrence relationships, the server establishes edges in the knowledge point graph.

[0133] An initial vector representation is assigned to each node. These vectors can be random or obtained through a pre-trained word embedding model (such as Word2Vec) to contain the initial semantic information of the vocabulary.

[0134] The server inputs the initial vector of each node into the embedding layer to obtain the target feature embedding data. Then, for each node in the knowledge point graph, the attention scores between it and all other nodes are calculated.

[0135] The attention scores are calculated by means of dot product, concatenation and then passing through a neural network. These scores reflect the degree of dependence between nodes. For example, the attention score of the "black pig" node to the "red spot" node may be relatively high because they often appear simultaneously in texts describing diseases.

[0136] The softmax function is used to normalize the attention scores to obtain the attention weights of each node to other nodes.

[0137] The server weighted-aggregates the neighbor node features of each node according to the attention weights to update the feature representation of the node.

[0138] For the "black pig" node, the server will weighted-aggregate the features of neighbor nodes such as "red spot" and "bacterial septicemia" according to its attention weights to other nodes to obtain the updated feature representation of the "black pig" node.

[0139] Perform a linear transformation and a non - linear activation (such as ReLU) on the aggregated feature representation to generate the target feature representation corresponding to each node.

[0140] For the entire reference aquaculture feedback text, the server integrates the target feature representations of all its corresponding nodes (such as "black pig", "red spot", "bacterial septicemia") (such as average pooling), and maps them to the basic aquaculture feedback semantic space through a fully - connected layer. The basic aquaculture feedback semantic space is a high - dimensional vector that contains the semantic information of the key knowledge points in the text and the association relationships between them.

[0141] In a possible implementation manner, the method further includes:

[0142] Step D110, obtain the second aquaculture feedback label attribute set, and input the second aquaculture feedback label attribute set into the basic aquaculture feedback mining network. The second aquaculture feedback label attribute set is generated according to the collaborative aquaculture feedback text sequence of the second sample abnormal aquaculture trigger event, and the basic aquaculture feedback mining network includes the set aquaculture feedback semantic space sequence.

[0143] In this embodiment, the server first selects a second sample abnormal aquaculture trigger event (such as "slow growth of live pigs"), and generates a second aquaculture feedback label attribute set from the collaborative aquaculture feedback text sequence of this event. The second aquaculture feedback label attribute set contains the key aquaculture feedback label attributes related to the second sample abnormal aquaculture trigger event.

[0144] Specifically, the server selects a representative second sample abnormal aquaculture trigger event from the historical data. According to this event, the server retrieves the collaborative aquaculture feedback text sequence within a period of time before and after the occurrence of the second sample abnormal aquaculture trigger event. Then, using natural language processing technology and the domain knowledge base, the server extracts the key aquaculture feedback label attributes from the collaborative aquaculture feedback text sequence and constructs the second aquaculture feedback label attribute set.

[0145] The server inputs the second aquaculture feedback label attribute set into the pre - constructed basic aquaculture feedback mining network. The basic aquaculture feedback mining network internally contains the set aquaculture feedback semantic space sequence, which is used to represent different types of aquaculture feedback scenarios.

[0146] For example, first, ensure that the basic aquaculture feedback mining network has been initialized and is in a trainable state. Then, load the second aquaculture feedback label attribute set into the input layer of the basic aquaculture feedback mining network in a certain format (such as a serialized label attribute vector).

[0147] Step D120: Using the basic aquaculture feedback mining network, in the set of preset aquaculture feedback semantic space sequences, retrieve the preset aquaculture feedback semantic spaces corresponding to each aquaculture feedback label attribute before the second label node in the second set of aquaculture feedback label attributes, perform semantic space feature collaboration on the retrieved preset aquaculture feedback semantic spaces to generate the estimated aquaculture feedback semantic space corresponding to the second label node, and based on the estimated aquaculture feedback semantic space corresponding to the second label node, generate the second estimated confidence corresponding to the aquaculture feedback label attribute on the second label node in the second set of aquaculture feedback label attributes. The second label node is determined from each aquaculture feedback label node in the second set of aquaculture feedback label attributes, and the second estimated confidence is used to reflect the estimated derivative metric value between the aquaculture feedback text corresponding to the second label node and the forward aquaculture feedback text sequence corresponding to the second label node.

[0148] In this embodiment, in the basic aquaculture feedback mining network, the server traverses each second label node in the second set of aquaculture feedback label attributes. For each second label node, the server retrieves the preset aquaculture feedback semantic spaces corresponding to all the label nodes before it and performs semantic space feature collaboration processing on these preset aquaculture feedback semantic spaces to generate the estimated aquaculture feedback semantic space corresponding to the second label node.

[0149] Specifically, for each second label node, the retrieval unit in the network searches for the corresponding preset aquaculture feedback semantic space among the label nodes before it. The collaboration unit performs feature collaboration processing on the retrieved multiple preset aquaculture feedback semantic spaces to capture the internal connections and temporal features between them. After the collaboration processing, the estimated aquaculture feedback semantic space corresponding to the second label node is generated.

[0150] Based on the estimated aquaculture feedback semantic space corresponding to the second label node, the server generates the second estimated confidence corresponding to the aquaculture feedback label attribute on this node, and this confidence reflects the estimated derivative metric value between this label attribute and its forward aquaculture feedback text sequence. For example, the prediction unit in the network uses the estimated aquaculture feedback semantic space and calculates the second estimated confidence through a specific algorithm (such as logistic regression, neural network output layer, etc.).

[0151] Step D130: Generate a second training error based on the second estimated confidence corresponding to each aquaculture feedback label attribute on each second label node in the second set of aquaculture feedback label attributes.

[0152] Step D140: Update the weight and bias information of the basic aquaculture feedback mining network according to the second training error until the second training termination requirement is met, and generate the target aquaculture feedback mining network.

[0153] In this embodiment, the server compares the second estimated confidence degrees on all second tag nodes in the second aquaculture feedback tag attribute set with a certain metric (such as cross-entropy loss) between the actual tag attributes to generate a second training error. Then, the server uses this second training error to update the weight and bias information of the basic aquaculture feedback mining network. Specifically, the second estimated confidence degrees can be compared with the actual tag attributes to calculate the second training error. Then, using optimization techniques such as the backpropagation algorithm, the weight and bias information of the network is updated according to the second training error. This process may involve multiple iterations until the training error is reduced below a certain threshold or a preset number of training rounds is reached.

[0154] After sufficient training iterations, the performance of the basic aquaculture feedback mining network reaches a stable state. At this time, the server regards it as the target aquaculture feedback mining network and uses it for subsequent real-time data processing and analysis tasks. During the training process, the performance of the network is regularly evaluated to ensure that its performance on the validation set meets expectations. Once the network meets the training termination requirements (such as the error being lower than the threshold, the training rounds being completed, etc.), the server saves it as the target aquaculture feedback mining network for subsequent use.

[0155] Through the above steps, the server trains the basic aquaculture feedback mining network and generates the target aquaculture feedback mining network. This process involves multiple links such as extracting the tag attribute set from the collaborative aquaculture feedback text sequence of the second sample abnormal aquaculture trigger event, inputting it into the network for training, generating the estimated semantic space and confidence degrees, calculating the training error, and updating the network weights and biases. The finally generated target network can more accurately understand and process the complex semantic associations and temporal features in the aquaculture feedback data, providing strong support for subsequent aquaculture anomaly detection, cause analysis and other tasks.

[0156] In a possible implementation manner, the basic aquaculture feedback mining network includes a retrieval unit, a collaboration unit, and a prediction unit. The retrieval unit is used to retrieve the set aquaculture feedback semantic space, the collaboration unit is used for semantic space feature collaboration, and the prediction unit is used to output the second estimated confidence degree.

[0157] Step D140 includes: updating the weight and bias information of the collaboration unit and the prediction unit in the basic aquaculture feedback mining network according to the second training error until the second training termination requirement is met, and generating the target aquaculture feedback mining network.

[0158] In this embodiment, the basic aquaculture feedback mining network is a complex deep learning model, which consists of three main parts: a retrieval unit, a collaboration unit, and a prediction unit. Each unit undertakes a specific task and jointly supports the overall function of the network.

[0159] Retrieval unit: Responsible for retrieving the semantic space related to the current processed label node from the set of aquaculture feedback semantic space sequences.

[0160] Collaboration unit: Performs feature collaboration processing on the retrieved multiple semantic spaces to capture the internal connections and temporal features between them.

[0161] Prediction unit: Based on the feature representation after collaboration processing, outputs the second estimated confidence level for the current processed label node.

[0162] During the training process, the server calculates the second training error based on the difference between the second estimated confidence level generated by the network and the actual label attributes. This error is an important indicator to measure the performance of the network and is used to guide subsequent weight and bias updates.

[0163] Specifically, the server compares the second estimated confidence level output by the prediction unit with the actual label attributes obtained through a certain method (such as manual annotation, historical data verification, etc.). Then, using a predefined loss function (such as cross-entropy loss, mean squared error, etc.), the server calculates the difference between the second estimated confidence level and the actual label attributes to obtain the second training error.

[0164] After obtaining the second training error, the server uses this second training error to update the weight and bias information of the collaboration unit and the prediction unit in the basic aquaculture feedback mining network. Since the retrieval unit is mainly responsible for data retrieval and does not involve complex calculation processes, its weights and biases usually remain unchanged during the training process.

[0165] Specifically, the server adopts the backpropagation algorithm to propagate the second training error layer by layer backward from the prediction unit to the collaboration unit. During the propagation process, the error is distributed to each weight and bias in each unit. According to the error gradient, the server calculates the adjustment amount for each weight and bias in the collaboration unit and the prediction unit. The calculation of the adjustment amount may involve the setting of hyperparameters such as the learning rate and momentum. The server applies the calculated adjustment amount to the weights and biases of the collaboration unit and the prediction unit to update the values of these parameters. The updated parameters will be used for the next round of forward propagation and error calculation.

[0166] The above error calculation, weight, and bias update processes are repeated for multiple iterations until the performance of the network reaches a preset standard or the training error is reduced to an acceptable range. Each iteration uses new training data to generate new second estimated confidence levels and training errors. After each iteration, the server evaluates the performance of the network (such as accuracy, recall, F1 score, etc.) and compares it with the preset training termination requirements. If the performance of the network meets the training termination requirements or the number of training rounds reaches the preset upper limit, the server stops the training process. Finally, the server saves the currently optimized basic aquaculture feedback mining network as the target aquaculture feedback mining network for subsequent real-time data processing and analysis tasks.

[0167] Through the above steps, the server optimizes the basic aquaculture feedback mining network and generates the target aquaculture feedback mining network. This process involves iteratively updating the weight and bias information of the collaborative unit and the prediction unit to minimize the second training error. As the iteration progresses, the performance of the network gradually improves and finally meets the preset training termination requirements. The generation of the target aquaculture feedback mining network not only improves the accuracy and efficiency of the intelligent monitoring system but also provides more reliable and useful data analysis support for aquaculture personnel.

[0168] In a possible implementation manner, step S140 includes:

[0169] Step S141, according to the text arrangement order of the global aquaculture feedback text sequence, converge each retrieved set aquaculture feedback semantic space to generate a set aquaculture feedback semantic space sequence.

[0170] Step S142, perform semantic space feature collaboration on the set aquaculture feedback semantic space sequence to generate a set of aquaculture feedback semantic collaboration spaces. The set of aquaculture feedback semantic collaboration spaces includes aquaculture feedback semantic collaboration spaces that match the number of texts in the global aquaculture feedback text sequence.

[0171] Step S143, from the set of aquaculture feedback semantic collaboration spaces, obtain the aquaculture feedback semantic collaboration space on the aquaculture feedback text label node of the target aquaculture feedback text as the target aquaculture feedback semantic collaboration space of the global aquaculture feedback text sequence.

[0172] In this embodiment, the server first orderly converges the previously retrieved respective preset aquaculture feedback semantic spaces according to the text arrangement order of the global aquaculture feedback text sequence to generate a preset aquaculture feedback semantic space sequence. Specifically, the server reviews the previous processing steps to ensure that the corresponding preset aquaculture feedback semantic space has been retrieved for each global aquaculture feedback text. According to the order of the texts in the global aquaculture feedback text sequence, the server arranges these preset aquaculture feedback semantic spaces in sequence to generate a preset aquaculture feedback semantic space sequence.

[0173] After obtaining the preset aquaculture feedback semantic space sequence, the server needs to further process these spaces using the feature collaboration technology to capture the internal connections and temporal features between them and generate a set of aquaculture feedback semantic collaboration spaces. Specifically, the server selects a suitable feature collaboration algorithm (such as graph attention network, long short-term memory network LSTM, etc.) to process the preset aquaculture feedback semantic space sequence. The selected feature collaboration algorithm is applied to iteratively process the preset aquaculture feedback semantic space sequence. During the processing, the feature representations in each preset aquaculture feedback semantic space and their dependencies can be considered to generate new feature representations, which not only contain the information in the original preset aquaculture feedback semantic space but also incorporate the relevant information in other preset aquaculture feedback semantic spaces. After multiple iterative processes, the server generates a set of aquaculture feedback semantic collaboration spaces. Each aquaculture feedback semantic collaboration space in this set corresponds to an aquaculture feedback text in the global aquaculture feedback text sequence and contains the comprehensive semantic information between this aquaculture feedback text and its preceding and following aquaculture feedback texts.

[0174] Finally, the server needs to extract the aquaculture feedback semantic collaboration space on the aquaculture feedback text label node of the target aquaculture feedback text from the set of aquaculture feedback semantic collaboration spaces as the target aquaculture feedback semantic collaboration space of the global aquaculture feedback text sequence. Specifically, the server first determines the position of the target aquaculture feedback text in the global aquaculture feedback text sequence, that is, its corresponding aquaculture feedback text label node. According to the position of the label node, the server finds the corresponding aquaculture feedback semantic collaboration space from the set of aquaculture feedback semantic collaboration spaces. This space not only contains the semantic information of the target aquaculture feedback text itself but also incorporates the comprehensive semantic information between its preceding and following texts, so it can more comprehensively reflect the context environment of the target aquaculture feedback text. The server saves the extracted target aquaculture feedback semantic collaboration space for subsequent tasks such as correlation analysis and anomaly detection.

[0175] Through the above steps, the server performs feature collaborative processing on each set breeding feedback semantic space in the global breeding feedback text sequence and generates a target breeding feedback semantic collaborative space. This process not only considers the temporal relationship between texts but also captures the internal connections and dependencies between them through feature collaborative technology. The finally generated target breeding feedback semantic collaborative space provides more comprehensive and accurate data support for subsequent tasks such as semantic analysis and anomaly detection.

[0176] Figure 2 FIG. shows a hardware structure diagram of a breeding feedback mining system 100 provided by an embodiment of the present application for implementing the above-mentioned artificial intelligence-based breeding feedback big data mining method, as Figure 2 shown, the breeding feedback mining system 100 may include a processor 110, a machine-readable storage medium 120, a bus 130, and a communication unit 140.

[0177] In a possible design, the breeding feedback mining system 100 may be a single server or a server group. The server group may be centralized or distributed (for example, the breeding feedback mining system 100 may be a distributed system). In some embodiments, the breeding feedback mining system 100 may be local or remote. For example, the breeding feedback mining system 100 may access information and / or data stored in the machine-readable storage medium 120 via a network. Again, for example, the breeding feedback mining system 100 may be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the breeding feedback mining system 100 may be implemented on a breeding feedback mining system. By way of example only, the breeding feedback mining system may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any aggregation thereof.

[0178] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions for the breeding feedback mining system 100 to execute or use to complete the exemplary methods described in the present application.

[0179] In a specific implementation process, one or more processors 110 execute computer-executable instructions stored in the machine-readable storage medium 120, so that the processors 110 can execute the artificial intelligence-based breeding feedback big data mining method of the above method embodiment. The processors 110, the machine-readable storage medium 120, and the communication unit 140 are connected through the bus 130, and the processors 110 may be used to control the transceiver actions of the communication unit 140.

[0180] For the specific implementation process of the processor 110, reference can be made to the various method embodiments executed by the above-mentioned aquaculture feedback mining system 100. Their implementation principles and technical effects are similar, and will not be elaborated here in this embodiment.

[0181] In addition, an embodiment of the present application also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned method for mining aquaculture feedback big data based on artificial intelligence is implemented.

[0182] It should be noted that, in order to simplify the description of the present application disclosure and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, sometimes multiple features are merged into one embodiment, drawing, or description thereof. Similarly, it should be noted that, in order to simplify the description of the present application disclosure and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. An artificial intelligence-based method for mining big data on breeding feedback, characterized in that The method includes: Obtaining a collaborative breeding feedback text sequence of a target breeding feedback text and a target abnormal breeding trigger event from big data of breeding feedback; Determining a derivative breeding feedback text sequence of the target breeding feedback text from the collaborative breeding feedback text sequence of the target abnormal breeding trigger event; Generating a global breeding feedback text sequence based on the derivative breeding feedback text sequence and the target breeding feedback text, and loading the global breeding feedback text sequence into a target breeding feedback mining network; the target breeding feedback mining network includes a set breeding feedback semantic space sequence, and the target breeding feedback mining network is generated by knowledge learning based on the collaborative breeding feedback text sequence of an example abnormal breeding trigger event. The set breeding feedback semantic space in the set breeding feedback semantic space sequence is the target breeding feedback semantic space of the breeding feedback text, and the target breeding feedback semantic space of the breeding feedback text is generated by strengthening the semantic association link of the basic breeding feedback semantic space of the breeding feedback text; Using the target breeding feedback mining network, retrieving the set breeding feedback semantic space corresponding to each breeding feedback text in the global breeding feedback text sequence in the set breeding feedback semantic space sequence, and performing semantic space feature collaboration on the retrieved set breeding feedback semantic spaces to generate a target breeding feedback semantic collaboration space of the global breeding feedback text sequence; Obtaining the breeding feedback semantic space of the target breeding feedback text and the abnormal breeding feature space of the target abnormal breeding trigger event, and loading the target breeding feedback semantic collaboration space, the breeding feedback semantic space of the target breeding feedback text, and the abnormal breeding feature space of the target abnormal breeding trigger event into a target pairing neural network to generate the correlation degree between the target breeding feedback text and the target abnormal breeding trigger event; The determining the derivative breeding feedback text sequence of the target breeding feedback text from the collaborative breeding feedback text sequence of the target abnormal breeding trigger event includes: For the collaborative breeding feedback text in the collaborative breeding feedback text sequence of the target breeding feedback text and the target abnormal breeding trigger event, extracting the target breeding feedback semantic space of the breeding feedback text; the target breeding feedback semantic space is generated by strengthening the semantic association link of the basic breeding feedback semantic space of the breeding feedback text; Determining multiple derivative breeding feedback texts of the target breeding feedback text from the collaborative breeding feedback text sequence of the target abnormal breeding trigger event according to the semantic correlation degree between the target breeding feedback text and the target breeding feedback semantic spaces of the collaborative breeding feedback texts, in descending order of the semantic correlation degree; Sorting each derivative breeding feedback text according to the collaborative time sequence position with the target abnormal breeding trigger event to generate a derivative breeding feedback text sequence of the target breeding feedback text.

2. The method for mining breeding feedback big data based on artificial intelligence according to claim 1, characterized in that, For the collaborative farming feedback text in the collaborative farming feedback text sequence of the target farming feedback text and the target abnormal farming trigger event, extracting the target farming feedback semantic space of the farming feedback text includes: Generating a current farming feedback label attribute set according to the farming feedback label attribute of the current farming feedback text, and loading the current farming feedback label attribute set into the target semantic association link reinforcement network; the current farming feedback text is the collaborative farming feedback text in the collaborative farming feedback text sequence of the target farming feedback text or the target abnormal farming trigger event, and the target semantic association link reinforcement network includes a basic farming feedback semantic space sequence; Using the target semantic association link reinforcement network, in the basic farming feedback semantic space sequence, retrieving the basic farming feedback semantic space corresponding to the farming feedback label attribute of the current farming feedback text, and strengthening the semantic association link of the retrieved basic farming feedback semantic space to generate the target farming feedback semantic space of the current farming feedback text.

3. The method for mining big data of aquaculture feedback based on artificial intelligence according to claim 2, characterized in that, The method further includes: Obtaining a first farming feedback label attribute set, and inputting the first farming feedback label attribute set into the basic semantic association link reinforcement network; the first farming feedback label attribute set is generated according to the collaborative farming feedback text sequence of the first sample abnormal farming trigger event, and the basic semantic association link reinforcement network includes the basic farming feedback semantic space sequence; Using the basic semantic association link reinforcement network, in the basic farming feedback semantic space sequence, retrieving the basic farming feedback semantic spaces corresponding to the farming feedback label attributes before the first label node in the first farming feedback label attribute set, strengthening the semantic association links of the retrieved basic farming feedback semantic spaces to generate the estimated farming feedback semantic space corresponding to the first label node, and generating the first estimated confidence corresponding to the farming feedback label attribute on the first label node in the first farming feedback label attribute set according to the estimated farming feedback semantic space corresponding to the first label node; the first label node is determined from each farming feedback label node of the first farming feedback label attribute set, and the first estimated confidence is used to reflect the estimated derivative metric value between the farming feedback text corresponding to the first label node and the forward farming feedback text sequence corresponding to the first label node; Generating a first training error according to the first estimated confidence corresponding to the farming feedback label attributes on each first label node in the first farming feedback label attribute set; Updating the weight and bias information of the basic semantic association link reinforcement network according to the first training error until the first training termination requirement is met, and generating the target semantic association link reinforcement network.

4. The method for mining big data of aquaculture feedback based on artificial intelligence according to claim 3, wherein, The generating the first estimated confidence corresponding to the farming feedback label attribute on the first label node in the first farming feedback label attribute set according to the estimated farming feedback semantic space corresponding to the first label node includes: Perform semantic feature domain conversion on the estimated aquaculture feedback semantic space corresponding to the first label node to generate the basic semantic feature domain vector distribution corresponding to the first label node; the basic semantic feature domain vector distribution includes semantic feature domain vectors corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence. Perform semantic dependence enhancement on the basic semantic feature domain vector distribution to generate the target semantic feature domain vector distribution corresponding to the first label node; the target semantic feature domain vector distribution includes the first estimated confidence levels corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence, and the reference aquaculture feedback text sequence includes aquaculture feedback texts corresponding to each aquaculture feedback label attribute in the first aquaculture feedback label attribute set. Determine the first estimated confidence level corresponding to the aquaculture feedback label attribute of the first aquaculture feedback label attribute set on the first label node from the target semantic feature domain vector distribution.

5. The method for mining big data of aquaculture feedback based on artificial intelligence according to claim 3, characterized in that, The basic semantic association link enhancement network includes a retrieval unit, a mapping unit, a semantic association link enhancement unit, and a prediction unit. The retrieval unit is used to retrieve the basic aquaculture feedback semantic space, the mapping unit is used for semantic association link mapping, the semantic association link enhancement unit is used for semantic association link enhancement, and the prediction unit is used to generate the first estimated confidence level. Updating the weights and bias information of the basic semantic association link enhancement network according to the first training error until the first training termination requirement is met to generate the target semantic association link enhancement network includes: Updating the weights and bias information of the mapping unit, the semantic association link enhancement unit, and the prediction unit in the basic semantic association link enhancement network according to the first training error until the first training termination requirement is met to generate the target semantic association link enhancement network.

6. The method for mining big data on aquaculture feedback based on artificial intelligence according to claim 2, wherein The method further includes: Obtain the text knowledge point distributions corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence. Through a graph self-attention processing network, perform graph self-attention semantic mining on the text knowledge point distributions to generate the basic aquaculture feedback semantic spaces corresponding to each reference aquaculture feedback text. Generate the basic aquaculture feedback semantic space sequence according to the basic aquaculture feedback semantic spaces corresponding to each reference aquaculture feedback text. Among them, the steps of obtaining the text knowledge point distributions corresponding to each reference aquaculture feedback text in the reference aquaculture feedback text sequence, and performing graph self-attention semantic mining on the text knowledge point distributions through a graph self-attention processing network to generate the basic aquaculture feedback semantic spaces corresponding to each reference aquaculture feedback text include: Perform word segmentation on each reference aquaculture feedback text in the reference aquaculture feedback text sequence, segment the continuous reference aquaculture feedback text into a word segmentation sequence, perform part-of-speech tagging on the word segmentation sequence, and then use named entity recognition technology to extract the key entity sequence in the word segmentation sequence. The key entities in the key entity sequence include aquaculture varieties, disease names, and symptom descriptions. Extract the core vocabulary in the key entity sequence as knowledge points through a keyword extraction algorithm. Take the extracted knowledge points as nodes, establish edges according to the co-occurrence relationship of the nodes in the corresponding reference breeding feedback text, construct a knowledge point graph, and assign an initial vector representation to each node in the knowledge point graph; Input the initial vector representation of each node into an embedding layer to obtain target feature embedding data. For each node in the knowledge point graph, calculate the attention score of this node to all other nodes. The attention score reflects the dependence degree of the current node on other nodes. The calculation of the attention score is realized through dot product, concatenation and then through a neural network, and the softmax function is used to normalize the attention score to obtain the attention weight of each node to other nodes; According to the attention weight, weight and aggregate the features of the neighbor nodes of each node to obtain the updated feature representation of the current node. After linear transformation and non-linear activation of the aggregated updated feature representation, generate the target feature representation corresponding to each node. For each reference breeding feedback text, integrate the target feature representations of its corresponding nodes, and map the integrated target feature representation to the basic breeding feedback semantic space through a fully connected layer.

7. The method for mining big data of aquaculture feedback based on artificial intelligence according to claim 1, wherein The method further includes: Obtain a second set of breeding feedback label attributes, and input the second set of breeding feedback label attributes into the basic breeding feedback mining network; the second set of breeding feedback label attributes is generated according to the collaborative breeding feedback text sequence of the second sample abnormal breeding trigger event, and the basic breeding feedback mining network includes the set of breeding feedback semantic space sequences; Using the basic breeding feedback mining network, in the set of breeding feedback semantic space sequences, retrieve the set of breeding feedback semantic spaces corresponding to each breeding feedback label attribute before the second label node in the second set of breeding feedback label attributes, perform semantic space feature collaboration on the retrieved set of breeding feedback semantic spaces, generate the estimated breeding feedback semantic space corresponding to the second label node, and generate the second estimated confidence corresponding to the breeding feedback label attribute on the second label node in the second set of breeding feedback label attributes according to the estimated breeding feedback semantic space corresponding to the second label node; the second label node is determined from each breeding feedback label node in the second set of breeding feedback label attributes, and the second estimated confidence is used to reflect the estimated derivative metric value between the breeding feedback text corresponding to the second label node and the forward breeding feedback text sequence corresponding to the second label node; Generate a second training error according to the second estimated confidence corresponding to each breeding feedback label attribute on each second label node in the second set of breeding feedback label attributes; Update the weight and bias information of the basic breeding feedback mining network according to the second training error until the second training termination requirement is met, and generate the target breeding feedback mining network; Among them, the basic aquaculture feedback mining network includes a retrieval unit, a collaboration unit, and a prediction unit. The retrieval unit is used to retrieve a set aquaculture feedback semantic space. The collaboration unit is used for semantic space feature collaboration. The prediction unit is used to output a second estimated confidence level. Updating the weights and bias information of the basic aquaculture feedback mining network according to the second training error until the second training termination requirement is met to generate the target aquaculture feedback mining network includes: Updating the weights and bias information of the collaboration unit and the prediction unit in the basic aquaculture feedback mining network according to the second training error until the second training termination requirement is met to generate the target aquaculture feedback mining network.

8. The method for mining breeding feedback big data based on artificial intelligence according to claim 1, wherein, Performing semantic space feature collaboration on each retrieved set aquaculture feedback semantic space to generate a target aquaculture feedback semantic collaboration space for the global aquaculture feedback text sequence, including: According to the text arrangement order of the global aquaculture feedback text sequence, converging each retrieved set aquaculture feedback semantic space to generate a set aquaculture feedback semantic space sequence; Performing semantic space feature collaboration on the set aquaculture feedback semantic space sequence to generate a set of aquaculture feedback semantic collaboration spaces; the set of aquaculture feedback semantic collaboration spaces includes aquaculture feedback semantic collaboration spaces that match the number of texts in the global aquaculture feedback text sequence; From the set of aquaculture feedback semantic collaboration spaces, obtaining the aquaculture feedback semantic collaboration space on the aquaculture feedback text label node of the target aquaculture feedback text as the target aquaculture feedback semantic collaboration space of the global aquaculture feedback text sequence.

9. A farming feedback mining system, characterized in that, The aquaculture feedback mining system includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions, or codes. The processor is used to execute the programs, instructions, or codes in the memory to implement the artificial intelligence-based aquaculture feedback big data mining method according to any one of claims 1-8 above.

Images