An intelligent retrieval and indexing method and system for fishery data

Abstract

The application relates to the field of fishery big data application, in particular to a fishery data intelligent retrieval and indexing method and system. The method comprises the following steps: collecting multi-source micro-ecological data, and performing collaborative calibration on the multi-source micro-ecological data to obtain calibrated micro-ecological data; identifying semantic events according to the calibrated micro-ecological data, and generating corresponding event abstracts; updating a semantic correlation knowledge representation structure according to the event abstracts, and performing causal correlation reasoning on a user query about the influence of environmental micro changes on fishery macro output based on the updated semantic correlation knowledge representation structure, and taking the updated semantic correlation knowledge representation structure as a data indexing basis of a vertical search engine to retrieve and respond to the query. The method solves the problems that the semantic correlation representation structure of the existing fishery vertical search engine faces a knowledge modeling bottleneck, the calculation pressure of incremental updating and relationship reasoning is large, and the timeliness and relevance of the query result are insufficient when processing micro-ecological data.

Description

Technical Field

[0001] This application relates to the field of big data applications in fisheries, specifically to a method and system for intelligent retrieval and indexing of fisheries data. Background Technology

[0002] In the application of big data in fisheries, vertical search engines typically rely on semantic association structures to connect multiple sources of macro data, such as fishing vessel catch records, marine environmental monitoring data, and aquatic product market transaction data, to establish macro relationships such as "fishing season - fish distribution" and "catch volume changes - market price fluctuations," providing overall decision support for management departments.

[0003] However, with the increasing complexity and sophistication of the marine ecological environment and the growing demand for predictive management, national standards require the integration of more detailed micro-ecological data, such as the content and particle size distribution of microplastics in water bodies, and the composition of microbial communities and their metabolites. This type of data is highly specialized and dimensional, and is mandated for integration into existing fisheries big data platforms, with the expectation of deep integration with existing macro-data to support more refined analysis.

[0004] Existing semantic association structures are mainly designed around macroscopic entities and direct relationships such as "fish, sea areas, and catch volume," making it difficult to support microscopic and biochemical concepts such as "microplastic particle size distribution" and "specific microbial community products." They also struggle to depict the complex and indirect causal chains between these concepts and factors like "feeding habits" and "individual growth and development." For example, it's difficult to represent and infer "decreased respiratory efficiency" or even "long-term changes in catch volume" from "accumulation of microplastics of specific particle sizes in fish gills," creating a gap in knowledge modeling between the macro and micro levels. Newly added microscopic ecological data is difficult to effectively integrate into the knowledge system.

[0005] Meanwhile, the data collection frequency for microplastics and microorganisms is extremely high, and the data volume is enormous. During incremental updates, the knowledge graph needs to frequently identify new entities, establish new relationships, and adjust relationship weights, leading to a sharp increase in memory and computational load. Index reconstruction and inference delays are significant, and the semantic structure struggles to reflect the latest marine ecological status in a timely manner. Due to update lag and insufficient correlation capabilities, when faced with complex, causal queries such as "the long-term impact of increased microplastic concentration in a certain sea area on the catch of a specific fish species," the system cannot provide timely and reliable results along the multi-layered inference path of "microplastic characteristics—plankton—food chain—catch," severely limiting its application value in refined management and predictive decision-making.

[0006] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0007] This application discloses an intelligent retrieval and indexing method and system for fishery data, aiming to solve the problems faced by existing fishery vertical search engines when processing micro-ecological data, such as the knowledge modeling bottleneck of semantic association representation structure, the high computational pressure of incremental updates and relational reasoning, and the insufficient timeliness and relevance of query results.

[0008] The technical solution of this application is as follows: Firstly, this application discloses an intelligent retrieval and indexing method for fishery data, including: Collect multi-source micro-ecological data and perform collaborative calibration on the multi-source micro-ecological data. Identify and correct abnormal data based on the correlation between each data source and the preset normal fluctuation range to obtain calibrated micro-ecological data. Based on the calibrated micro-ecological data, semantic events are identified and corresponding event summaries are generated. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics, and confidence information. The semantic association knowledge representation structure is updated based on the event summary. Nodes and relation edges representing the first relationship between semantic events and entities related to fishery macro-output are created or updated within the semantic association knowledge representation structure. Based on the updated semantic association knowledge representation structure, causal association inference is performed on user queries about the impact of micro-environmental changes on fishery macro-output. This serves as the data index basis for the vertical search engine to retrieve and respond to queries. The first relationship represents the influence relationship between related entities and their semantic events based on the time sequence.

[0009] Furthermore, before performing collaborative calibration on multi-source micro-ecological data, the method also includes: Acquiring underwater acoustic signals; Extracting acoustic features from underwater acoustic signals; Identify the type of systematic environmental interference based on acoustic characteristics; The multi-source micro-ecological data are compensated and calibrated according to the type of systemic environmental disturbance to obtain the compensated and calibrated multi-source micro-ecological data. Collaborative verification of multi-source micro-ecological data after compensation and calibration.

[0010] Furthermore, based on the calibrated micro-ecological data, semantic events are identified, and corresponding event summaries are generated, which also includes: Acquire multispectral fluorescence data; Calculate the microalgal bloom intensity index based on multispectral fluorescence data; The optical interference compensation amount and ecological risk adjustment factor were determined based on the microalgal bloom intensity index, and the benchmark event triggering threshold used to identify abnormal increases in microplastic concentration were dynamically adjusted to obtain the dynamic event triggering threshold. Based on the dynamic event triggering threshold, events of abnormal increase in microplastic concentration are identified from the calibrated microecological data. The abnormal increase in microplastic concentration is a type of semantic event. The dynamic event triggering threshold, microalgal bloom intensity index and optical interference compensation amount are recorded in the corresponding event summary.

[0011] Furthermore, based on the calibrated micro-ecological data, semantic events are identified, and corresponding event summaries are generated, including: Collect data on fishery activities; Obtain historical fisheries production data and records of historical ecological events; Based on fishery activity data, historical fishery production data, and historical ecological event records, the identification rules and thresholds for semantic events are dynamically adjusted. Based on the adjusted recognition rules and thresholds, semantic events are identified from the calibrated micro-ecological data; Generate corresponding event summaries based on the identified semantic events, and record the adjusted recognition rules and thresholds in the event summaries.

[0012] Furthermore, a corresponding event summary is generated based on the identified semantic events, and the adjusted recognition rules and thresholds are recorded in the event summary, including: The adjusted recognition rules and thresholds are structured and encoded, mapping the parameters of rule type, adjustment range and applicable conditions in the recognition rules and thresholds to predefined short codes or enumeration values; A preset summary template is selected based on the type and importance of the identified semantic events. The summary template includes fields for representing the core information of the event and specific fields for carrying the structured encoding recognition rules and thresholds. The structured recognition rules and thresholds are filled into the selected summary template to obtain the filled event summary; The padded event digest is compressed to generate a compact event digest for transmission.

[0013] Furthermore, the adjusted recognition rules and thresholds are structured and encoded, including: Construct a parameter association graph based on the semantic relationships between the parameters in the adjusted recognition rules and thresholds; Identify core parameter clusters and associated paths from the parameter association graph; The core parameter cluster and associated path are mapped to predefined combined short codes or hierarchical enumeration values ​​to obtain the encoding result; Based on the preset semantic consistency judgment rules, the encoding results are verified to ensure that the encoding results are semantically consistent with the core parameter cluster and associated paths.

[0014] Furthermore, based on preset semantic consistency judgment rules, the encoding results are verified, including: Periodically obtain the latest fisheries policies, marine environmental early warning information, and historical event patterns; The semantic consistency judgment rules are dynamically adjusted based on the latest fisheries policies, marine environmental early warning information, and historical event patterns. The encoding results are verified according to the dynamically adjusted semantic consistency judgment rules.

[0015] Furthermore, the structured encoded recognition rules and thresholds are filled into the selected summary template to obtain the filled event summary, which also includes: Based on the identified semantic event type, obtain the corresponding preset macroeconomic impact assessment rules; The macroeconomic impact of the identified semantic events is assessed based on the adjusted identification rules and thresholds, combined with the macroeconomic impact assessment rules, to obtain preliminary macroeconomic impact assessment results. The preliminary macroeconomic impact assessment results are compared with the pre-acquired macroeconomic knowledge graph or historical reasoning results to identify whether there are any conflicts and obtain conflict identification results. Based on the conflict identification results, conflict markers are applied to the adjusted identification rules and thresholds. Add conflict markers and preliminary macroeconomic impact assessment results to the filled event summary.

[0016] Furthermore, based on the type of the identified semantic event, corresponding preset macroeconomic impact assessment rules are obtained, including: Get current geographic location information; Get current season information; Obtain current fisheries activity data; Based on the identified semantic event type, current geographical location information, current season information, and current fishery activity data, macroeconomic impact assessment rules are matched from a pre-set rule base.

[0017] Secondly, this application also discloses a fisheries data intelligent retrieval and indexing system, including: The data acquisition and calibration module is used to acquire multi-source micro-ecological data and perform collaborative calibration on the multi-source micro-ecological data. It identifies and corrects abnormal data based on the correlation between each data source and the preset normal fluctuation range to obtain calibrated micro-ecological data. The identification module is used to identify semantic events based on the calibrated micro-ecological data and generate corresponding event summaries. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics and confidence information. The reasoning module is used to update the semantic association knowledge representation structure based on the event summary. In the semantic association knowledge representation structure, it creates or updates nodes and relation edges that represent the first relationship between semantic events and entities related to the macro-output of fisheries. Based on the updated semantic association knowledge representation structure, it performs causal association reasoning on user queries about the impact of micro-environmental changes on the macro-output of fisheries. It also serves as the data index basis for the vertical search engine to retrieve and respond to queries. The first relationship is used to represent the influence relationship between related entities and their semantic events based on the time sequence.

[0018] Beneficial Effects: This application can identify semantic events and generate structured event summaries, transforming complex micro-changes into understandable and processable information. It overcomes the limitations of existing technologies where micro-concepts are difficult to effectively integrate into the knowledge system. Furthermore, by updating the semantic association knowledge representation structure and creating or updating the first relation nodes and edges between micro-events and macro-output in fisheries, this application achieves multi-level causal reasoning from micro-changes to macro-impacts, effectively solving the technical problems of knowledge modeling bottlenecks and the inability to perform deep reasoning in existing technologies. Accordingly, this application can serve as the data indexing foundation for vertical search engines, efficiently retrieving and responding to user queries regarding the impact of environmental micro-changes on macro-output in fisheries. This significantly enhances the application value of fisheries vertical search engines in refined management and prediction, overcoming the shortcomings of insufficient timeliness and relevance of query results in existing technologies. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a data indexing method for a vertical search engine provided in this application.

[0020] Figure 2 A flowchart of a vertical search engine data indexing system provided in this application.

[0021] In the diagram: 1. Acquisition and calibration module; 2. Identification module; 3. Inference module. Detailed Implementation

[0022] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0023] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] In the application of big data in fisheries, building an efficient and accurate vertical search engine is fundamental to supporting intelligent management and scientific research in fisheries. Traditional fisheries vertical search engines rely on semantic association structures to connect multi-source information such as fishing vessel records, marine environmental monitoring data (water temperature, salinity, etc.), and aquatic product market transaction data. This allows for the establishment of macro-level relationships such as "fishing season - fish distribution area" and "catch changes - market price fluctuations," providing management departments with trend-based and macro-level decision support.

[0025] However, with the increasing complexity of the marine ecological environment and the growing demand for refined and predictive management, existing technologies have revealed significant shortcomings. On the one hand, newly introduced microscopic ecological data (such as microplastic particle size distribution, specific microbial communities and their metabolites) belong to the microscopic and biochemical level, which differs in granularity and hierarchy from the traditional conceptual system built around macroscopic entities such as "fish, sea areas, and catch volumes." Existing semantic structures struggle to express and link the multi-level causal chains between "accumulation of microplastics of specific particle sizes in fish gills," "decreased respiratory efficiency," and "long-term changes in catch volumes." New data is difficult to integrate into existing knowledge graphs, and its potential value remains unrealized.

[0026] On the other hand, microplastic and microbial data are collected at extremely high frequencies and in massive volumes. Each update may trigger partial or overall adjustments to the knowledge graph, including new entity identification, relationship establishment, and weight adjustments, resulting in a heavy memory and computational burden. Under current hardware conditions, high-frequency updates lead to a significant increase in index reconstruction and inference latency, making it difficult for the knowledge representation structure to reflect the latest marine ecological status in a timely manner. Under the dual constraints of insufficient semantic expression and lagging updates, when researchers and managers raise deep-seated, causal queries such as "the long-term impact of increased microplastic concentration in a specific sea area on the local grouper catch," traditional search engines are unable to perform multi-layered inference along the path of "microplastic characteristics—plankton—food chain—catch," and are also unable to integrate the latest microscopic data in a timely manner, making it difficult to provide timely and highly relevant results, thus weakening their application value in refined management and predictive decision-making.

[0027] Reference Figure 1 In response, this application proposes an intelligent retrieval and indexing method for fisheries data, including: S1000: Collects multi-source micro-ecological data and performs collaborative calibration on the multi-source micro-ecological data. Based on the correlation between each data source and the preset normal fluctuation range, it identifies and corrects abnormal data to obtain calibrated micro-ecological data. S2000: Based on the calibrated micro-ecological data, semantic events are identified and corresponding event summaries are generated. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics and confidence information. S3000: Update the semantic association knowledge representation structure based on the event summary. Create or update nodes and relation edges in the semantic association knowledge representation structure to represent semantic events and their first relationship with entities related to fishery macro-output. Based on the updated semantic association knowledge representation structure, perform causal association reasoning on user queries about the impact of micro-environmental changes on fishery macro-output. Use this as the data index basis for the vertical search engine to retrieve and respond to queries.

[0028] The first relation is used to represent the influence relationship between related entities and their semantic events based on the time sequence.

[0029] Specifically, in this embodiment, multi-source micro-ecological data refers to data collected from different sensors, monitoring devices, or data platforms that reflect the micro-level state of marine or freshwater ecosystems. This data may include, but is not limited to: water temperature, salinity, dissolved oxygen, pH value, chlorophyll a concentration, plankton species and quantity, microplastic content, specific microbial community composition, nutrient concentrations (such as nitrates and phosphates), etc., and typically exhibits high frequency, high dimensionality, and heterogeneity.

[0030] Collaborative calibration refers to the unified processing of micro-ecological data from different data sources to eliminate or reduce data deviations caused by sensor differences, environmental interference, transmission errors, etc. Collaborative calibration may include steps such as data cleaning, missing value imputation, outlier detection and correction, and data fusion, with the aim of improving the overall consistency and accuracy of the data.

[0031] Semantic events refer to events identified in micro-ecological data that have specific meanings and impacts, such as microalgal blooms, abnormally high microplastic concentrations, and abnormal proliferation of specific pathogenic bacterial communities. These events are usually indicated by the coordinated changes of a series of micro-indicators and may have potential impacts on macro-level fisheries output.

[0032] An event summary is a structured description of identified semantic events, including event type (e.g., microalgal bloom), occurrence time, duration, key statistical information (e.g., peak and average chlorophyll a concentration during the bloom), and confidence information (indicating the reliability of the event identification). This structured data facilitates subsequent storage, retrieval, and inference.

[0033] Semantic association knowledge representation structure is a graph structure used to store and represent knowledge in the fisheries field. It contains entities (such as fish, sea areas, microplastics, microorganisms, catch, etc.) and primary relations, attribute relations, composition relations, etc. between entities. In this application, this structure particularly emphasizes the causal relationship between micro-ecological events and macro-output of fisheries.

[0034] Causal reasoning refers to inferring the potential impact or cause of known micro-ecological events on macro-fisheries output based on semantic association knowledge representation structures, using logical reasoning or machine learning models. For example, an increase in microplastic concentration in a specific sea area can lead to the inference that fish growth in that area may slow down, further affecting catch volume.

[0035] The intelligent retrieval and indexing method for fisheries data in this application firstly involves the collaborative calibration of multi-source micro-ecological data after collection. For example, data such as water temperature, salinity, dissolved oxygen, chlorophyll a concentration, and microplastic content can be acquired from different types of sensors, including water quality monitoring buoys, remote sensing satellites, and underwater robots. To address measurement biases, noise interference, or transmission errors, Z-score normalization or IQR (interquartile range) methods can be used to identify anomalous data points, which are then corrected using linear interpolation, spline interpolation, or prediction models based on historical data. Another approach is to establish a multi-sensor fusion model, such as based on Kalman filtering or extended Kalman filtering, to fuse similar data from different sensors to obtain more accurate estimates. For example, for water temperature data, both buoy sensor and satellite remote sensing data can be received simultaneously, and the differences between the two can be calibrated using a fusion algorithm, correcting anomalous readings caused by local environmental factors.

[0036] In some embodiments, collaborative calibration can be implemented according to the following steps. First, for each monitoring indicator in the multi-source micro-ecological data, historical data bins are established according to the combination of "sea area, season, depth range, and indicator type". For each bin, based on historical monitoring data from the past few years (e.g., 3-5 years), the statistical mean μ, standard deviation σ, first quartile Q1, third quartile Q3, and interquartile range IQR = Q3−Q1 are calculated respectively, thereby constructing the preset normal fluctuation range of the indicator in the corresponding spatiotemporal context. For example, when the real-time observed value x of a certain monitoring indicator satisfies |x−μ| / σ≤2 and is located within the interval [Q1−1.5·IQR, Q3+1.5·IQR], it can be considered to be within the normal fluctuation range. For indicators exhibiting strong seasonal variations (such as water temperature and chlorophyll a concentration), a sliding time window (e.g., the most recent 30 or 60 days) can be used to dynamically update μ, σ, Q1, Q3, and IQR to avoid situations where long-term historical distributions are no longer adapted to the current environmental conditions. For observation points falling outside the aforementioned statistically normal range, the system combines historically labeled ecological anomaly events with normal samples to assess the probability of them falling at the tail of the anomaly distribution. Based on this, it distinguishes between "statistically extreme values" and "points that may represent true ecological anomalies," thus avoiding mistaking true anomalies for noise during subsequent anomaly correction.

[0037] Secondly, for the same type of indicator from different data sources (such as water quality monitoring buoys, remote sensing satellites, underwater robots, etc.), a horizontal comparative analysis is performed within a preset time alignment window. Specifically, the correlation coefficient r and average deviation Δx between each data source can be calculated, for example, by comparing the differences and correlations between buoy water temperature and satellite-retrieved water temperature. When multiple data sources show a uniform overall shift in the same direction over a longer time window, and the correlation coefficient is high while the changes in individual sensors are not isolated, this shift can be identified as a systematic bias rather than a single sensor failure or random noise.

[0038] Furthermore, for detected single-point outliers, interpolation methods or predictive models can be used for correction. For example, a reasonable value at that moment can be estimated using linear interpolation, spline interpolation, or a predictive model based on historical time series (such as the ARIMA model or a simple recurrent neural network model), and the estimated value can replace the original outlier. For identified systematic biases, a multi-sensor fusion model can be established for correction. In one optional implementation, a state-space model based on Kalman filtering or extended Kalman filtering can be constructed. Similar indicators from different sensors with different observation resolutions and accuracies can be used as observation vectors, and the real environmental state can be used as the implicit state. The optimal estimate of the indicator at the corresponding time and spatial location is obtained through a filtering process and used as the output value after collaborative calibration. For example, for water temperature data, buoy measurements, satellite inversion values, and local measurements from underwater robots can be input into a Kalman filter. Through state update and observation update stages, the bias of each data source is estimated and corrected, and the fused and calibrated water temperature estimate is output.

[0039] Secondly, semantic events are identified based on the calibrated micro-ecological data, and corresponding event summaries are generated. For example, by continuously monitoring the calibrated chlorophyll a concentration, if it exceeds a preset threshold within a short period (e.g., within 24 hours) and persists for a period (e.g., 48 hours), it can be identified as a microalgal bloom event. Event summaries can be generated using a template-filling method: a preset microalgal bloom event summary template is provided, containing key statistical information such as event type, occurrence time, duration, key indicators (e.g., peak and average chlorophyll a concentration), and confidence level. Once a bloom event is identified, the corresponding data is filled into this template to form a structured event summary. For example: Event type: microalgal bloom; Occurrence time: 2023-10-26, 08:00; Duration: 72 hours; Key indicator statistics: peak chlorophyll a concentration 150 μg / L, average concentration 80 μg / L; Confidence level: 0.95.

[0040] In some embodiments, the identification of semantic events can be based on a formalized set of event identification rules. Each semantic event type corresponds to a set of rules, which includes several triggering conditions, logical relationships between conditions, duration requirements, and confidence calculation methods. Each triggering condition can be expressed as "a certain monitoring indicator meets a specific threshold constraint within a given time window," such as "chlorophyll a concentration continuously exceeds 50 μg / L within the last 24 hours" or "dissolved oxygen decreases by more than 30% compared to the historical average within the last 48 hours." Multiple conditions can be combined through logical relationships (such as AND and OR) and additional duration constraints can be added, such as requiring certain conditions to be met simultaneously for at least 48 hours before an event is triggered.

[0041] For each semantic event, the initial baseline event trigger threshold can be determined based on the following three aspects: First, according to the threshold range specified in national or local ecological environment, water quality, or fisheries standards, for example, using the critical value of a certain indicator reaching a certain level as a candidate threshold for event triggering; Second, through statistical analysis of historical ecological event samples, for example, using past microalgal bloom events as positive samples and normal periods as negative samples, calculating evaluation indicators based on the sensitivity and specificity under different thresholds, thereby selecting a threshold that can take into account both the false negative and false positive rates; Third, fine-tuning the above statistical results by domain experts based on experience to ensure that the threshold is reasonable for the region, the target species, and the management objectives. For example, for the "abnormal increase in microplastic concentration event," historical data on the impact of different microplastic concentration levels on fisheries output can be statistically analyzed, and an ROC curve can be used to select a concentration value with optimal overall performance (such as 500 particles / liter) as the baseline event trigger threshold.

[0042] For ease of consistent terminology, the baseline event triggering threshold used to identify a semantic event is denoted as T_base (or simply baseline event triggering threshold), and the dynamically adjusted event triggering threshold is denoted as T_dyn (or simply dynamic threshold). Unless otherwise specified, these notations will be used throughout the text. Based on this, the baseline event triggering threshold can be dynamically adjusted using multispectral fluorescence data and fisheries activity data. In one implementation, the microalgal bloom intensity index is first calculated based on multispectral fluorescence data. This index can be obtained by normalizing the fluorescence intensity ratio or spectral integral value of a specific excitation / emission wavelength channel to a value in the range of 0 to 1, with a higher value indicating a stronger bloom. Then, the optical interference compensation amount C_optics is calculated based on a preset functional relationship between this index and optical interference. Simultaneously, the ecological risk adjustment factor R_eco is calculated based on the empirical or model relationship between this index and ecological risk. For example, when the bloom intensity index is close to 1, C_optics and R_eco take higher values.

[0043] Subsequently, the baseline event trigger threshold T_base, used to identify "abnormal increases in microplastic concentration events," is dynamically adjusted to obtain the dynamic event trigger threshold T_dyn. In a simple linear adjustment example, T_dyn can be set as T_base + C_optics − k·R_eco·T_base, where k is a preset risk amplification coefficient used to control the degree to which ecological risk lowers the threshold. For example, when the baseline threshold is 500 particles / L, multispectral fluorescence data indicates the presence of a moderate microalgal bloom, the calculated optical interference compensation C_optics is 50 particles / L, and the ecological risk adjustment factor R_eco corresponds to k·R_eco·T_base of 0, then the dynamic event trigger threshold can be adjusted to 550 particles / L. At this point, when the microplastic concentration in the calibrated microecological data reaches 560 particles / L, it can be identified as a genuine "abnormal increase in microplastic concentration event" based on T_dyn, while simultaneously avoiding misjudging false increases caused solely by optical interference as events.

[0044] In practical applications, for different types of semantic events, the dynamic event triggering threshold T_dyn can be adjusted based solely on the optical interference compensation amount C_optics and the ecological risk adjustment factor R_eco, or solely on the fisheries risk score R_fishery, or a combination of the above factors can be considered within the same framework. Those skilled in the art can modify the calculation formula of T_dyn by equivalent transformations, factor additions or subtractions, or weight adjustments according to the target sea area, target species, and management objectives. As long as it conforms to the basic idea of ​​"adaptively adjusting the threshold based on environmental optical interference and fisheries risk on the basis of the baseline event triggering threshold T_base," it is considered to fall within the protection scope of this invention.

[0045] For dynamic adjustments based on fisheries activity data and historical fisheries production data, a fisheries risk score, R_fishery, can be defined to comprehensively reflect factors such as the intensity of current fisheries activities and the frequency of historical disease or yield reduction events. For example, standardized indicators such as stocking density, catch per unit time, and frequency of similar historical ecological events can be weighted and summed to obtain an R_fishery value in the range of 0 to 1. Based on this, the baseline event triggering threshold T_base for a certain semantic event can be dynamically adjusted according to T_dyn = T_base·(1−γ·R_fishery), where γ is a preset risk sensitivity coefficient. When R_fishery is large, the dynamic event triggering threshold T_dyn decreases accordingly, thereby increasing sensitivity to potentially high-risk events.

[0046] Taking shrimp farming disease risk as an example, historical data shows that during the high-density shrimp farming season in a certain sea area, the incidence of shrimp diseases increases significantly when the water temperature exceeds 28℃. In the absence of significant fishery activity or historical risks, 28℃ can be used as a baseline threshold for abnormally high water temperature events. However, in the current scenario of high-density shrimp farming and frequent historical disease records, the dynamic threshold can be adjusted to 27.5℃ using R_fishery calculation. When the calibrated micro-ecological data monitors a water temperature of 27.6℃ for 24 hours, even if the original baseline threshold has not been reached from a general environmental perspective, the method of this application can still identify it as a semantic event of "potential shrimp disease risk," and record the dynamic threshold of 27.5℃ and the corresponding risk score in the event summary, improving the timeliness and targeting of the early warning.

[0047] Finally, the semantic association knowledge representation structure is updated based on the event summary, and causal association inference is performed based on the updated structure. This structure is also used as the data index foundation for the vertical search engine to retrieve and respond to queries. For example, after identifying the event summary of a microalgal bloom event, the system checks whether nodes and relationships related to this event type already exist in the semantic association knowledge representation structure. If not, new nodes (e.g., microalgal bloom event_ID123) and relationship edges (e.g., causing _hypoxia_affecting_fish growth) are created. If they already exist, the attributes of the relevant nodes or relationships are updated, such as increasing the number of events or updating the impact intensity. Furthermore, microalgal bloom event nodes can be connected to macroscopic output entities such as decreased dissolved oxygen, increased fish mortality, and reduced catches in the graph through first relationship edges.

[0048] When a user queries the impact of microalgal blooms on grouper catch, the reasoning module traverses the semantic association knowledge representation structure, starting from the microalgal bloom event node, along the first relation edge, through intermediate nodes such as decreased dissolved oxygen, difficulty in grouper breathing, and increased grouper mortality, and finally deduces the impact on the reduction of grouper catch. These causal paths and results are used as the index basis to respond to queries.

[0049] In some embodiments, the semantic association knowledge representation structure can be formalized as a directed weighted graph G=(V,E). The node set V includes, but is not limited to, nodes representing micro-ecological semantic events (e.g., "microalgal bloom event", "abnormal increase in microplastic concentration event"), environmental states (e.g., "decline in dissolved oxygen", "decrease in water transparency"), macro-fisheries outputs (e.g., "reduction in grouper catch", "increase in shrimp disease incidence"), and policy or management measures. The edge set E represents the first relationship between nodes, with each edge containing attributes such as relationship type, direction, strength weight, confidence level (conf), and optional time delay parameter.

[0050] The first relationship edge can be obtained from three types of information sources. The first is a rule base provided by domain experts, which formalizes expert knowledge into causal rules of "several micro-events and environmental conditions → macro-impact," such as "a moderate to severe algal bloom event occurring for three consecutive days usually leads to a significant decrease in dissolved oxygen in the sea area." The second is statistical learning results based on historical data. Granger causality tests, time-series correlation analysis, or regression models can be applied to historical monitoring sequences and event records. When a micro-event has a statistically significant leading effect on a macro-indicator, a first relationship edge pointing from that micro-event to the macro-indicator can be added to the graph, and the correlation strength or regression coefficient can be mapped to the initial weight of the edge. The third is the output of ecological simulation models, such as "the impact of changes in a certain type of plankton on the population size of a specific fish species" simulated based on trophic level kinetic models or individual baseline models. This can also be converted into the weight and time delay attribute of the first relationship edge.

[0051] During operation, as new event summaries are continuously input, the weights and confidence levels of existing first-relation edges can be incrementally updated. For example, when a summary of a "microalgal bloom event" matches a subsequently observed "dissolved oxygen decrease" summary in both time and space, and aligns with the expected direction of the existing first-relation "microalgal bloom event → dissolved oxygen decrease," the weights and confidence levels of that causal edge can be strengthened. Conversely, if the expected dissolved oxygen decrease is not observed after multiple occurrences of "microalgal bloom events," the weights of that causal edge can be appropriately reduced. Simple update methods can employ exponentially weighted moving averages or Bayesian updates. For instance, each hit can be considered as obs_effect=1, and each miss as obs_effect=0, updated according to w_new=(1−η)·w_old+η·obs_effect, where η is the learning rate parameter. Confidence levels can be calculated by combining historical sample size, hit rate, and statistical significance.

[0052] When a user initiates a query about the impact of micro-environmental changes on macro-fisheries output, causal reasoning can be performed based on the aforementioned semantic association knowledge representation structure. Specifically, the system first locates the corresponding starting node (e.g., "abnormal increase in microplastic concentration in a specific sea area") and target macro-output node (e.g., "change in local grouper catch") in the graph based on the query keywords. Then, within a limited path length range (e.g., no more than four hops), it searches for all directed or partially undirected causal paths from the starting point to the ending point, and calculates a path score for each path. This score can be a combination of the product of the weights of each edge on the path and the average or minimum confidence level. When there are multiple paths with different intermediate mechanisms (e.g., through the "plankton feeding obstruction" path and the "decreased fish respiration efficiency" path), the system can sort the scores of different paths, select several paths with higher scores as the main reasoning chains, and output the intermediate nodes as part of the reasoning explanation.

[0053] When dealing with uncertainty and contradictory evidence, the total score of the path sets corresponding to different macroscopic conclusions (e.g., "increased catch", "decreased catch", "no significant impact") can be calculated separately, normalized to similar probability values, and each potential conclusion and its confidence interval can be given in the query results. When certain causal edges or event summaries are marked as having conflict risks, a discount factor can be applied to the paths passing through these edges in the path score, or the relevant reasoning chains can be marked "requires verification" in the final output, thereby ensuring real-time response while alerting decision-makers to information uncertainty.

[0054] Those skilled in the art will understand that the path search and scoring process based on directed weighted graphs described above is only one preferred method for realizing causal reasoning. While adhering to the basic technical principle of "using the first relation edge in the semantic association knowledge representation structure to perform path-level reasoning and uncertainty handling on the relationship between micro-ecological semantic events and macro-fishery outputs," random walks, graph neural network-based reasoning models, or other probabilistic graphical models can also be used.

[0055] An example of this embodiment is as follows: When a user queries "the long-term impact of increased microplastic concentration in a specific sea area on local grouper catch," existing systems typically fail to provide satisfactory results because they lack the ability to deduce the impact of microplastics on plankton from their physicochemical properties, then deduce the impact on the grouper food chain from changes in plankton, and finally assess the potential impact on catch. The method in this application, however, relies on a semantic association knowledge representation structure to reason along the causal chain of "increased microplastic concentration" → "impaired plankton feeding" → "reduced food sources for grouper" → "slower grouper growth / increased mortality" → "reduced catch," thereby providing more accurate and insightful query results. This better supports refined fisheries management and forecasting, provides decision-makers with scientific and comprehensive information support, and significantly enhances the value of big data applications in fisheries.

[0056] In another embodiment of this application, the following steps are further proposed before performing collaborative calibration on multi-source micro-ecological data: S1001: Acquires underwater acoustic signals; S1002: Extracting acoustic features from underwater acoustic signals; S1003: Identify the type of systematic environmental interference based on acoustic characteristics; S1004: Compensate and calibrate the multi-source micro-ecological data according to the type of systematic environmental disturbance to obtain the compensated and calibrated multi-source micro-ecological data; S1005: Collaborative verification of multi-source micro-ecological data after compensation and calibration.

[0057] Specifically, underwater acoustic signal acquisition refers to the continuous or periodic acquisition of sound wave data in the water body through acoustic sensors or hydrophones deployed underwater. Underwater acoustic signals include both natural sounds (such as ocean waves, raindrops, and sounds emitted by marine life) and man-made sounds (such as ship navigation, fishing operations, and sonar detection), which are used to provide the raw data basis for subsequent identification of environmental interference.

[0058] Extracting acoustic features from underwater acoustic signals involves preprocessing and feature engineering the acquired raw acoustic signals, converting the raw waveform into analyzable feature vectors. Acoustic features can include time-domain features (such as root mean square, peak value, and zero-crossing rate), frequency-domain features (such as spectrum, Mel-frequency cepstral coefficients (MFCC), and energy band), and time-frequency-domain features (such as short-time Fourier transform (STFT) and wavelet transform). Through this feature extraction, the complex raw acoustic data is quantified into structured information, making it suitable for pattern recognition analysis by machine learning models or rule-based expert systems.

[0059] Identifying systematic environmental interference types based on acoustic features refers to using pattern recognition methods to classify and judge extracted acoustic features in order to distinguish different types of external interference sources. Specifically, machine learning algorithms such as Support Vector Machines (SVM), Neural Networks (NN), and deep learning models, or rule-based expert systems, can be used to analyze acoustic features and identify systematic environmental interference types such as ship noise within a specific frequency range, periodic sonar signals, and vocalization patterns of specific biological groups.

[0060] Compensation and calibration of multi-source micro-ecological data based on the type of systematic environmental interference yields compensated and calibrated multi-source micro-ecological data. This involves identifying specific types of environmental interference and then specifically correcting the micro-ecological data affected by that interference. For example, when sonar interference at a specific frequency is identified, sensor data such as temperature, salinity, and dissolved oxygen that may be affected by sound waves during the same period can be filtered, or compensated based on historical data and physical models. Compensation and calibration methods can include, but are not limited to, signal filtering, noise cancellation algorithms, and bias correction based on statistical models. Through this step, the impact of known systematic interference on the original micro-ecological data is eliminated or weakened as much as possible before entering collaborative calibration, thereby improving the purity of the data.

[0061] Collaborative verification of multi-source micro-ecological data after compensation calibration refers to reconfirming the compensation effect by utilizing redundant information and mutual corroboration relationships among multiple data sources after initial compensation calibration. For example, the consistency of identical or related indicator data collected by different sensors at the same time and in the same area can be compared, or the compensated data can be compared with historical trends and physical model predictions to verify the effectiveness of the compensation calibration. Through collaborative verification, it can be confirmed that the compensation process has not introduced new systematic errors, further improving data reliability and providing higher-quality input for subsequent collaborative calibration.

[0062] This application's solution introduces underwater acoustic signal acquisition, acoustic feature extraction, and systematic environmental interference type identification before collaborative calibration of multi-source micro-ecological data. This effectively identifies and quantifies the potential impact of external factors such as ship noise and sonar interference on micro-ecological data acquisition. Based on this, targeted compensation calibration is then performed, and the compensation effect is verified through collaborative validation. This achieves pre-emptive reduction of known systematic environmental interference, reducing the interference of environmental noise on data quality at its source. This avoids misjudging abnormal fluctuations caused by environmental interference as ecological anomalies, providing a purer and more reliable micro-ecological data foundation for subsequent collaborative calibration, semantic event recognition, and causal correlation inference.

[0063] In some preferred embodiments, this application can be specifically implemented as follows: Multiple sensors are deployed in a certain aquaculture area to collect multi-source micro-ecological data such as water temperature, salinity, dissolved oxygen, and pH value; simultaneously, hydrophones are deployed in the area to collect underwater acoustic signals. When a fishing boat passes by or conducts sonar detection, the hydrophones collect the corresponding acoustic signals. The system extracts spectral features, MFCC, and other acoustic features from the acoustic signals, and identifies systemic environmental interference types such as "ship noise" or "sonar pulses" using a pre-trained machine learning model (e.g., a model based on a convolutional neural network CNN). Once "ship noise" is identified, the system compensates and calibrates the simultaneously collected micro-ecological data (e.g., dissolved oxygen sensor readings that may be affected by sound waves) according to the frequency and intensity of the noise, such as by using bandpass filtering or correction methods based on historical noise models. After compensation, the system compares the compensated dissolved oxygen data with data from other sensors in the same area that are not directly affected by acoustic interference, or performs trend analysis with historical data from the same period, to collaboratively verify the effectiveness of the compensation calibration. For example, the compensation calibration is considered effective when the calibrated dissolved oxygen data is consistent with other sensor data or historical trends. The data obtained after the above processing will more realistically reflect the micro-ecological environment, providing more accurate input for subsequent identification of semantic events such as red tides and hypoxia events, and ultimately supporting vertical search engines to perform more reliable causal reasoning.

[0064] In another embodiment of this application, S2000 further includes: S2100: Acquires multispectral fluorescence data; S2110: Calculate the microalgal bloom intensity index based on multispectral fluorescence data; S2120: The optical interference compensation amount and ecological risk adjustment factor are determined based on the microalgal bloom intensity index, and the benchmark event triggering threshold used to identify abnormal increases in microplastic concentration is dynamically adjusted to obtain the dynamic event triggering threshold. S2130: Based on the dynamic event triggering threshold, identify microplastic concentration abnormal increase events from the calibrated microecological data. The microplastic concentration abnormal increase event is a type of semantic event. Record the dynamic event triggering threshold, microalgal bloom intensity index and optical interference compensation amount in the corresponding event summary.

[0065] Specifically, collecting multispectral fluorescence data refers to acquiring real-time fluorescence spectral information of the water body at different excitation and emission wavelengths using multispectral fluorescence sensors deployed in the water. This multispectral fluorescence data reflects the concentration of photosynthetic pigments such as chlorophyll a and phycobiliproteins in the water, thereby indicating the biomass and physiological state of microalgae.

[0066] The calculation of the microalgal bloom intensity index based on multispectral fluorescence data involves processing the collected multispectral fluorescence data using algorithms, such as using fluorescence intensity ratios or spectral integral values ​​in specific bands to quantify the severity of microalgal blooms in the water. The microalgal bloom intensity index can intuitively reflect the occurrence, development, and intensity changes of blooms.

[0067] In practical applications, determining the optical interference compensation and ecological risk adjustment factor based on the microalgal bloom intensity index means that when the microalgal bloom intensity index reaches a certain level, the optical interference compensation amount is calculated based on a preset model or empirical relationship between the microalgal bloom intensity index and optical interference to correct the measured value of microplastic concentration. This is because strong algal blooms reduce water transparency, interfering with optical sensors used for microplastic detection. Furthermore, high-intensity microalgal blooms may indicate a deterioration of the aquatic ecosystem, increasing the risk of microplastic impacts on fishery ecosystems. Therefore, an ecological risk adjustment factor can be determined based on the microalgal bloom intensity index and historical data to weight the risk during event identification.

[0068] Furthermore, the baseline event trigger threshold used to identify abnormal increases in microplastic concentration is dynamically adjusted to obtain the dynamic event trigger threshold. This refers to applying the optical interference compensation amount and ecological risk adjustment factor to a preset initial or baseline event trigger threshold used to determine whether microplastic concentration has increased abnormally, after determining these factors. For example, in the presence of significant optical interference, the baseline event trigger threshold for microplastic concentration can be appropriately increased to avoid false alarms; when the ecological risk is high, the baseline event trigger threshold can be appropriately decreased to improve early warning sensitivity. The dynamic event trigger threshold obtained after the above adjustments can better adapt to real-time environmental changes.

[0069] Therefore, identifying abnormal increases in microplastic concentration from calibrated microecological data based on a dynamic event trigger threshold involves comparing the microplastic concentration value in the aforementioned compensated and calibrated microecological data with the dynamic event trigger threshold. When the microplastic concentration value exceeds this threshold, an abnormal increase in microplastic concentration is determined to have occurred. This abnormal increase in microplastic concentration is considered a type of semantic event.

[0070] Finally, the dynamic event trigger threshold, microalgal bloom intensity index, and optical interference compensation amount are recorded in the corresponding event summary. This means that the aforementioned dynamic event trigger threshold, microalgal bloom intensity index, and optical interference compensation amount are written as structured fields into the event summary, stored together with the event type, occurrence time, duration, key indicator statistics, and confidence level information. This provides more complete contextual information for subsequent causal association reasoning and facilitates the tracing and verification of the event identification process.

[0071] This application's solution, by introducing multispectral fluorescence data, enables real-time and accurate monitoring of microalgal blooms in water bodies, quantifying their intensity as a microalgal bloom intensity index. Based on this, it precisely calculates the optical interference compensation and ecological risk adjustment factor under the current environment, and dynamically adjusts the benchmark event trigger threshold used to identify abnormal increases in microplastic concentration. This threshold adapts to changes in the water's optical properties and ecological risk level. With this dynamic adjustment mechanism, the identification of abnormal increases in microplastic concentration no longer relies on a fixed static threshold, effectively avoiding misjudgments caused by optical interference such as microalgal blooms. Furthermore, it allows for flexible adjustment of the warning sensitivity based on actual ecological risk, thus more accurately and robustly identifying genuine abnormal increases in microplastic concentration. This provides high-quality event data for subsequent updates to the semantic association knowledge representation structure and causal reasoning.

[0072] In some preferred embodiments, it is assumed that sensors for monitoring microplastic concentration and multispectral fluorescence sensors are deployed in a certain aquaculture area. When the system detects an upward trend in microplastic concentration, it simultaneously analyzes multispectral fluorescence data. For example, if multispectral fluorescence data shows that a moderate microalgal bloom is occurring in the area, the system calculates an optical interference compensation based on a preset model and determines an ecological risk adjustment factor by combining the microalgal bloom intensity index and historical data. Assuming that the baseline event trigger threshold for identifying abnormal increases in microplastic concentration is 500 particles / L, this threshold is dynamically adjusted to 550 particles / L after considering the optical interference compensation and the ecological risk adjustment factor. At this point, if the microplastic concentration in the calibrated microecological data reaches 560 particles / L, the system will accurately identify this as an abnormal increase in microplastic concentration. In the generated event summary, in addition to the event type, occurrence time, duration, key indicator statistics and confidence information, the dynamic event trigger threshold of 550 algae / L used in the event identification, the microalgal bloom intensity index at that time (e.g., 0.7), and the optical interference compensation amount (e.g., 50 algae / L) will also be recorded.

[0073] In another embodiment of this application, S2000 is further proposed to include: S2200: Collects fisheries activity data; S2210: Obtain historical fishery production data and historical ecological event records; S2220: Dynamically adjust the recognition rules and thresholds for semantic events based on fishery activity data, historical fishery production data, and historical ecological event records; S2230: Identify semantic events from calibrated micro-ecological data based on the adjusted identification rules and thresholds; S2240: Generate corresponding event summaries based on the identified semantic events, and record the adjusted identification rules and thresholds in the event summaries.

[0074] Specifically, collecting fisheries activity data refers to gathering various types of information related to fisheries production, such as fishing operation type, fishing area, fishing time, number of fishing vessels, aquaculture density, and feed input. Fisheries activity data can originate from AIS (Automatic Identification System) data from fishing vessels, regulatory records from fisheries authorities, production logs from aquaculture farms, or IoT sensor data, providing real-time, directly relevant background information for semantic event identification. Acquiring historical fisheries production data and historical ecological event records involves retrieving historical catch volumes, aquaculture yields, disease occurrence records, red tide or algal bloom records, extreme weather events, etc., from databases, archives, or knowledge bases to obtain long-term trends and patterns, helping the system understand the potential meanings of specific micro-ecological changes in different contexts.

[0075] In practical applications, the identification rules and thresholds for semantic events can be dynamically adjusted based on fishery activity data, historical fishery production data, and historical ecological event records. Machine learning models or expert systems can be used to optimize the parameters of preset semantic event identification models (such as anomaly detection algorithms and pattern matching rules) in real time. For example, in specific fishing seasons or high-density aquaculture areas, when historical data shows that slight fluctuations in certain micro-indicators often correspond to higher risks, the system will automatically tighten the identification thresholds for relevant semantic events; while in non-production seasons or low-risk areas, the corresponding thresholds can be appropriately relaxed to prevent the system from becoming oversensitive.

[0076] Furthermore, based on dynamically adjusted identification rules and thresholds, semantic events are identified from calibrated micro-ecological data. This means the system no longer uses a uniform, fixed judgment standard, but instead employs dynamic standards that match the current fisheries environment and historical context to determine the existence of semantic events. For example, a 0.5℃ increase in water temperature is generally not considered abnormal, but if it occurs during the breeding season of a specific aquaculture species and historical records show that this temperature change has caused disease, the system will identify this change as a "potential disease risk event" based on the adjusted identification rules. After identification, a corresponding event summary is generated based on the identified semantic events, recording the adjusted identification rules and thresholds in the event summary. This ensures that the event summary clearly reflects the contextualized standards upon which the event identification is based. For example, the event summary could state "Identification rule: Abnormal increase in water temperature (threshold dynamically adjusted to +0.5℃, due to the breeding season of XX aquaculture species)," enhancing the transparency and interpretability of the event summary and enabling the subsequent causal reasoning module to accurately understand the basis for event identification.

[0077] The proposed solution constructs a dynamic adjustment mechanism for semantic event recognition by introducing fishery activity data, historical fishery production data, and historical ecological event records, so that the recognition of semantic events is no longer limited to static analysis of calibrated micro-ecological data.

[0078] In some preferred embodiments, the specific process can be as follows: Assume that in a certain sea area, the system continuously collects micro-ecological data such as water temperature, salinity, and dissolved oxygen, and simultaneously collects fishery activity data for that sea area, such as detecting a large amount of shrimp farming activity. The system finds from historical fishery production data that, in the past few years, under the same season and similar farming density conditions, the incidence of shrimp diseases significantly increases when the water temperature exceeds 28℃. Simultaneously, historical ecological event records show that this sea area has experienced several large-scale shrimp mortality events caused by specific pathogens, and these events are often accompanied by a rapid rise in water temperature. Based on the above information, the system dynamically adjusts the identification rules and thresholds for the semantic event of "abnormally high water temperature": In the absence of fishery activity or relevant historical disease records, a water temperature increase from 27℃ to 28℃ may not constitute a semantic event; however, given the current high-density shrimp farming and historical data showing that 28℃ is a high-risk threshold for disease, the system sets "water temperature exceeding 27.5℃ for 24 hours" as a new, more stringent identification threshold. When the calibrated micro-ecological data detects a water temperature of 27.6℃ for 24 hours, even if no alarm is triggered under the traditional static threshold, the proposed solution will still identify it as a semantic event of "potential shrimp disease risk". The subsequently generated event summary will include basic information such as event type and occurrence time, as well as adjusted identification rules and threshold information such as "Identification rule: water temperature exceeds 27.5℃ for 24 hours (dynamically adjusted, reason: high-density shrimp farming, historical disease risk)".

[0079] In another embodiment of this application, S2240 specifically includes: S2241: Perform structured encoding on the adjusted recognition rules and thresholds, mapping the parameters of rule type, adjustment range and applicable conditions in the recognition rules and thresholds to predefined short codes or enumeration values; S2242: Select a preset summary template based on the type and importance of the identified semantic events. The summary template includes fields for representing the core information of the event and specific fields for carrying the structured encoding recognition rules and thresholds. S2243: Fill the selected summary template with the structured encoded recognition rules and thresholds to obtain the filled event summary; S2244: Compress the padded event digest to generate a compact event digest for transmission.

[0080] Specifically, structured encoding of the adjusted identification rules and thresholds refers to converting key information in these rules and thresholds, such as rule type, adjustment range, and applicable conditions, into a standardized, machine-readable format. This can be achieved by mapping the above parameters to predefined short codes, enumeration values, or specific data structures, thereby ensuring the uniformity and parsability of the representation of rules and thresholds in the event summary. For example, the complex rule "When the water temperature exceeds 25 degrees Celsius and dissolved oxygen is below 3 mg / L, the threshold for the abnormal increase in microplastic concentration event is increased by 10%" can be encoded as "RULE_TYPE:TEMP_DO_ANOMALY_ADJ; ADJ_MAG:10%; COND:TEMP>25&DO<3".

[0081] The selection of pre-defined summary templates based on the type and importance of identified semantic events refers to the system choosing the summary template with the highest matching degree from a predefined template library based on different types of semantic events (such as abnormally high microplastic concentration events, red tide events, water pollution events, etc.) and their importance in the fisheries field. Each summary template contains standard fields for representing the core information of the event (such as event type, occurrence time, duration, key indicator statistics, and confidence information), and reserves specific fields to carry the structured and encoded identification rules and thresholds. Through this template-based design, different event summaries maintain structural uniformity, facilitating subsequent parsing, display, and indexing.

[0082] In practical applications, filling the selected summary template with the structured encoded recognition rules and thresholds to obtain the filled event summary means filling in the aforementioned structured encoding results according to the preset fields and their positions in the summary template, so that the generated event summary simultaneously contains the core information of the event itself and its dynamic recognition logic. Thus, the system can form a complete, structured event summary that records both the event characteristics and clearly records the adjusted recognition rules and thresholds used to identify the event.

[0083] Furthermore, the padded event digest is compressed to generate a compact event digest for transmission. This involves compressing the event digest using standard data compression algorithms (such as Gzip, Deflate, etc.) after its construction to reduce data size, lower network bandwidth usage, accelerate transmission, and reduce storage space consumption. In this way, the transmission and storage efficiency of the event digest in the system is improved without altering its semantic content.

[0084] The solution proposed in this application effectively solves the problems of unstructured data, data redundancy, and low transmission efficiency that may occur when recording adjusted identification rules and thresholds in event summaries by introducing structured coding, summary template selection and filling, and compression processing.

[0085] As a specific implementation, assume the system identifies an "abnormal increase in microplastic concentration event." According to this scheme, firstly, the adjusted identification rules and thresholds used to identify this event (e.g., "when water turbidity (NTU) exceeds 10 and dissolved oxygen (DO) is below 5 mg / L, the baseline event trigger threshold for the abnormal increase in microplastic concentration event is increased by 15%) are structured and encoded. The "rule type" is encoded as "MP_ANOMALY_ADJ", the "adjustment magnitude" is encoded as "+15%", and the "applicable conditions" are encoded as "TURBIDITY>10&DO<5", and mapped to predefined short codes or enumeration values ​​to form a compact encoded string. Subsequently, based on the type of the "abnormal increase in microplastic concentration event" and its importance in the fisheries sector, the system selects a summary template specifically for this type of event from the template library. This template may contain core fields such as "event ID", "occurrence time", "geographical location", and "peak microplastic concentration", and reserves a field named "ADJ_RULES_THRES" to carry the encoded identification rules and thresholds. Next, the structured, encoded identification rules and thresholds (e.g., the encoded string "MP_ANOMALY_ADJ:+15%;TURBIDITY>10&DO<5") are filled into the "ADJ_RULES_THRES" field to generate a complete, structured event summary. This allows for the recording of detailed information about the abnormal increase in microplastic concentration and its dynamic identification logic. Finally, the filled event summary is compressed, for example, using the Gzip algorithm to generate a compact event summary, which is then used as the data indexing basis for the vertical search engine. When a user queries "the impact of abnormally high microplastic concentration on fishery output," the search engine can quickly and accurately retrieve the event and its dynamic identification rules, and based on this, perform causal reasoning to output search results that are more relevant to the actual ecological and production context.

[0086] Furthermore, when performing structured encoding on the adjusted recognition rules and thresholds, the following methods can be used.

[0087] A1: Construct a parameter association graph based on the semantic relationships between the parameters in the adjusted recognition rules and thresholds; A2: Identify core parameter clusters and associated paths from the parameter association diagram; A3: Map the core parameter cluster and associated paths to predefined combined short codes or hierarchical enumeration values ​​to obtain the encoding result; A4: Based on the preset semantic consistency judgment rules, the encoding result is verified to ensure that the encoding result is semantically consistent with the core parameter cluster and associated path.

[0088] Specifically, a parameter association graph can be understood as a graphical representation used to depict the semantic relationships between different parameters in identification rules and thresholds. These parameters can include rule type, adjustment range, applicable conditions, etc. By representing parameters as nodes and their dependencies or constraints as edges in the parameter association graph, the inherent logical relationships and interdependencies between parameters are revealed. The core parameter cluster refers to a set of parameters that are highly correlated or share a common function in the parameter association graph; the association path represents the logical connection or data flow between core parameter clusters or individual parameters, reflecting the semantic chain of "which conditions and parameters work together to derive what adjustment result." Identifying core parameter clusters and association paths helps to extract the key components and operational mechanisms of identification rules and thresholds from complex parameter networks.

[0089] Furthermore, combined short codes or hierarchical enumeration values ​​are predefined, compact representations with specific meanings used to encode core parameter clusters and their associated paths. By mapping the identified core parameter clusters and associated paths to combined short codes or hierarchical enumeration values, structured, standardized, and efficient encoding of complex rules and thresholds can be achieved. For example, a specific combination of rule type and adjustment range can be assigned a unique short code, which can then be used to uniquely identify the meaning of the combination during subsequent storage and transmission. In practical applications, semantic consistency judgment rules are a set of pre-defined logical conditions or algorithms used to check whether the encoding results accurately reflect the semantics of the original core parameter clusters and associated paths, such as whether key parameters are omitted, whether mutually exclusive parameters are incorrectly combined, or whether pre-defined semantic constraints are violated. By verifying the encoding results, the semantic accuracy and reliability of the encoding can be guaranteed, avoiding deviations in subsequent reasoning or retrieval due to encoding errors.

[0090] The proposed solution first constructs a parameter association graph based on the semantic relationships between parameters in the adjusted recognition rules and thresholds. This transforms the recognition rules and thresholds from simple lists of isolated parameters into a structured graph representation of their semantic dependencies. Subsequently, core parameter clusters and association paths are identified from the parameter association graph, converging numerous scattered parameter relationships into a few representative key combinations and paths, providing focused and precise input for subsequent encoding. Building upon this, the core parameter clusters and association paths are mapped to predefined short codes or hierarchical enumeration values, unifying complex rules and thresholds into compact and standardized codes. This facilitates efficient processing in data indexing, transmission, and storage within vertical search engines and provides a clear and consistent parsing interface for upper-layer applications. Finally, the encoding results are validated according to predefined semantic consistency judgment rules. This allows for timely detection of semantic inconsistencies or combination errors during the encoding stage, preventing erroneous encoding from entering the semantic association knowledge representation structure and ensuring the correctness of data indexing and causal reasoning from the source.

[0091] In another embodiment of this application, A4 is further proposed to include: A41: Periodically obtain the latest fisheries policies, marine environmental early warning information, and historical event patterns; A42: Dynamically adjust semantic consistency judgment rules based on the latest fisheries policies, marine environmental early warning information, and historical event patterns; A43: Verify the encoding results according to the dynamically adjusted semantic consistency judgment rules.

[0092] Specifically, periodically acquiring the latest fisheries policies, marine environmental early warning information, and historical event patterns means that the system automatically collects relevant information from official channels, environmental monitoring agencies, and historical databases at preset time intervals. The latest fisheries policies may include fisheries management regulations issued by national or local governments, notices of closed fishing seasons, and adjustments to fishing quotas; marine environmental early warning information may include warnings of red tides, green tides, abnormal water temperatures, and pollutant exceedances; historical event patterns refer to typical ecological events and abnormal fisheries production events that have occurred in the past, as well as their evolution patterns. By continuously acquiring this information, real-time and comprehensive external evidence is provided for the subsequent dynamic adjustment of semantic consistency judgment rules.

[0093] Furthermore, dynamically adjusting semantic consistency judgment rules based on the latest fisheries policies, marine environmental early warning information, and historical event patterns means that the system does not use a fixed set of semantic consistency judgment rules, but rather a set of semantic consistency judgment rules that are updated and optimized in conjunction with the latest external information. For example, when a new fisheries policy is introduced, the encoding and verification logic of the semantic event identification rules and thresholds related to the policy will be adjusted synchronously to ensure that the encoding results comply with the latest policy constraints; when marine environmental early warning information such as red tides and abnormal water temperatures is issued, the semantic consistency judgment conditions of the ecological indicators related to these early warnings in the encoding verification can be tightened or relaxed accordingly to more accurately reflect the reasonable correlation patterns between parameters under the current environment; when historical event patterns indicate that the manifestation or impact path of a certain type of event changes under a specific background, the semantic consistency judgment rules will also update their verification thresholds or constraints accordingly.

[0094] In some embodiments, the semantic consistency judgment rule set can be stored in the form of a rule table or a rule base. Each rule entry specifies the applicable policy version, the applicable sea area, the applicable semantic event type, and the constraints on the values ​​of each parameter in the coding result. For example, a rule may stipulate that during a specific fishing ban period, the identification rules and threshold codes related to fishing intensity should not indicate the meaning of "allowing an increase in fishing volume," otherwise it is considered inconsistent with the policy.

[0095] To enable semantic consistency judgment rules to adapt to changes in fisheries policies and environmental conditions, periodic tasks can be set up to retrieve the latest fisheries policies, marine environmental early warning information, and historical event patterns from the fisheries management department's policy database, marine environmental monitoring and early warning system, and historical event analysis system. When a policy version update is detected, the environmental status of a certain sea area changes from normal to warning status, or a historical event pattern shows a new and significant trend, the system will modify or add relevant entries in the rule set according to a predefined update strategy. For example, when a sea area is added as a year-round fishing ban area, all fishing intensity enhancement rules related to that sea area will be marked as invalid or replaced with stricter constraints; when a certain type of red tide event occurs frequently recently, the threshold consistency requirements related to this type of event can be appropriately tightened to improve the sensitivity to high-risk coding results.

[0096] When verifying specific encoding results, the system first selects a subset of currently effective semantic consistency judgment rules based on the regional, time, and event type information contained in the encoding result. Then, it compares each parameter field in the encoding result (e.g., rule type short code, adjustment range, applicable condition combination, etc.) with the constraints in the rule subset one by one. If the values ​​of all key fields fall within the allowed range, the encoding result is determined to be semantically consistent with the current rule set. If any field value violates the constraints of one or more rule entries (e.g., encoding an increase in fishing intensity during a closed fishing season), the encoding result is marked as "semantically inconsistent," and the ID of the violated rule entry or a brief explanation is recorded. Encoding results marked as inconsistent can have their weight reduced or be prompted for manual review during subsequent event summary generation, macro-impact assessment, or causal inference processes, thereby improving the overall security and reliability of the system.

[0097] Therefore, verifying the encoding results based on the dynamically adjusted semantic consistency judgment rules means that after the system completes the structured encoding of the recognition rules and thresholds, it does not directly write the encoded results into the semantic association knowledge representation structure. Instead, it first compares and verifies the encoded results with the current version of the semantic consistency judgment rules. The goal of the verification is to comprehensively consider the latest fisheries policies, marine environmental early warning information, and historical event patterns to check whether the encoding results are semantically consistent with the core parameter clusters and association paths, and whether they are consistent with current policy requirements and environmental background, thereby improving the accuracy and reliability of the encoding semantic verification.

[0098] This application's solution introduces a step of "periodically acquiring the latest fisheries policies, marine environmental early warning information, and historical event patterns," enabling the semantic consistency judgment rules to be dynamically updated. This avoids the shortcomings of traditional static rules, which cannot adapt to adjustments in fisheries management policies, environmental emergencies, and long-term ecological pattern evolution. External information is continuously integrated into the adjustment process of the semantic consistency judgment rules, ensuring that the rule set can promptly reflect the actual operational status of the fisheries sector and preventing erroneous verification or releases due to rule lag.

[0099] In some preferred embodiments, it is assumed that in a specific sea area, the government issues a new fishing ban policy, prohibiting the fishing of certain fish species during specific months each year; simultaneously, the marine environmental monitoring department issues a red tide warning for that sea area, indicating that the red tide may cause abnormal fluctuations in multiple micro-ecological indicators; furthermore, the system, through historical event pattern analysis, finds that in similar red tide events in the past, misjudgments are prone to occur if the identification threshold for abnormally high microplastic concentrations is not dynamically adjusted. In this case, the method of this application periodically acquires the aforementioned new fishing ban policy, red tide warning information, and related historical event patterns, and dynamically adjusts the semantic consistency judgment rules accordingly: for example, updating the semantic event encoding verification rules related to fishing activities to comply with the new fishing ban requirements; and adjusting the semantic consistency judgment conditions of the threshold encoding verification rules related to abnormally high microplastic concentrations based on red tide warnings and historical patterns to enhance the verification of whether the abnormal threshold settings are reasonable. Ultimately, the system uses dynamically adjusted semantic consistency judgment rules to verify the structured encoding recognition rules and thresholds. Only when the encoding results are semantically consistent with the latest fisheries policies, environmental conditions, and historical experience will they be written into the subsequent data index and reasoning process, thereby ensuring the reliability and accuracy of the overall system in complex and ever-changing environments.

[0100] In another embodiment of this application, it is further proposed that, after S2243, the following is also included: S2243-1: Obtain the corresponding preset macroeconomic impact assessment rules based on the type of the identified semantic events; S2243-2: Based on the adjusted identification rules and thresholds, combined with the macro-impact assessment rules, assess the macro-impact of the identified semantic events to obtain preliminary macro-impact assessment results; S2243-3: Compare the preliminary macroeconomic impact assessment results with the pre-acquired macroeconomic knowledge graph or historical reasoning results to identify whether there are conflicts and obtain conflict identification results; S2243-4: Based on the conflict identification results, mark the adjusted identification rules and thresholds for conflict identification; S2243-5: Add conflict markers and preliminary macroeconomic impact assessment results to the filled event summary.

[0101] Specifically, after obtaining the populated event summary, the system first needs to retrieve the corresponding macroeconomic impact assessment rules from a pre-defined rule base based on the type of semantic event identified. For example, when the semantic event is "an abnormal increase in microplastic concentration," the system retrieves the macroeconomic impact assessment rules for this type of event from the rule base. These rules can provide an impact model on "the impact of increased microplastic concentration on the growth, catch, or stability of specific fish species or marine ecosystems."

[0102] In some embodiments, macroeconomic impact assessment rules can be stored in a rule base as structured rule entries. Each macroeconomic impact assessment rule typically includes two parts: a condition section and a conclusion section. The condition section may include information such as semantic event type, applicable geographical area, applicable season or time period, fishery activity pattern (e.g., aquaculture season, fishing season, closed season), and value ranges of key microeconomic indicators. The conclusion section describes the expected impact on specific fishery macroeconomic outputs (e.g., catch, average growth rate, juvenile survival rate, etc. of a certain economic fish species) when the above conditions are met. It can be expressed as a percentage range or interval, accompanied by a confidence level or credibility rating.

[0103] For example, a macroeconomic impact assessment rule for "anomaly in microplastic concentration events" can be stated as follows: "When the event type is anomaly in microplastic concentration, the geographical area is a North Atlantic fishing ground, the season is spring, and the microplastic concentration is in the range of 1.5 to 2.0 times the baseline value and lasts for more than 7 days, it is expected that the average growth rate of cod in the area will decrease by 5 to 8%, the survival rate of juvenile fish will decrease by 3 to 5%, and the confidence level is 0.8." Another rule for "anomaly in microplastic concentration events in South China Sea aquaculture areas during summer" can be stated as follows: Under high temperature conditions, the toxicity of microplastics is enhanced, which may lead to a decrease in grouper production of about 20% and increase the risk of certain diseases.

[0104] Once the system retrieves macroeconomic impact assessment rules from the rule base that match current geographic location information, seasonal information, and fisheries activity data based on identified semantic events and their event summaries, it can apply these rules to quantitatively or semi-quantitatively assess the potential macroeconomic impact of the event, obtaining a preliminary macroeconomic impact assessment result. Subsequently, the system compares this preliminary assessment result with a pre-constructed macroeconomic knowledge graph or historical reasoning results. For historical impact data in the macroeconomic knowledge graph corresponding to the event type, region, and season, its historical mean (Hist_mean) and standard deviation (Hist_std) can be calculated. The difference between the current preliminary assessment result (Pred) and historical statistical results is measured, for example, by calculating the deviation rate = |Pred−Hist_mean| / Hist_std. When the deviation rate exceeds a preset threshold (e.g., 2 or 3), or when the direction of the preliminary assessment result is opposite to historical conclusions, a macroeconomic-level conflict can be identified.

[0105] Once a conflict is identified, the system can mark it as a conflict based on the identification rules and thresholds used for this event identification, according to the degree of conflict. Conflict marking can include simple Boolean flags (e.g., "Conflict Present / No Conflict"), multi-level risk levels (e.g., "Low Conflict Risk," "Medium Conflict Risk," "High Conflict Risk"), or text labels with brief explanations (e.g., "Predicted impact intensity is significantly higher than the historical average; expert review recommended"). This conflict marking will be appended to the populated event summary along with the preliminary macroeconomic impact assessment results. This ensures that the event summary, as the basis for data indexing, includes both detailed descriptions of the micro-event and macroeconomic impact predictions, while also clearly indicating consistency or inconsistency with existing macroeconomic knowledge. This facilitates uncertainty management in subsequent reasoning and decision-making processes.

[0106] Furthermore, by combining the adjusted identification rules and thresholds with the acquired macroeconomic impact assessment rules, the potential macroeconomic impacts of the identified semantic events are evaluated, thus obtaining preliminary macroeconomic impact assessment results. For example, when the adjusted identification rules and thresholds indicate a significant increase in microplastic concentration over a prolonged period, the system, in conjunction with the macroeconomic impact assessment rules, can assess that the catch of specific economic fish species in the area may decline significantly in the future, or that there may be a significant negative impact on the stability of the corresponding ecosystem.

[0107] Subsequently, the preliminary macroeconomic impact assessment results are compared with pre-acquired macroeconomic knowledge graphs or historical inference results. The macroeconomic knowledge graph can include common sense in the fisheries field, expert knowledge, policies and regulations, and known macroeconomic first relationships between different ecological events; historical inference results can be macroeconomic impact analysis conclusions based on similar past events. Through comparison, inconsistencies or conflicts between the preliminary assessment results and existing macroeconomic knowledge or historical experience can be identified. For example, if the preliminary assessment predicts a "significant decline in catch," while the macroeconomic knowledge graph indicates that historically, under similar pollution levels, seasons, and fisheries activity backgrounds, catches typically only fluctuate slightly, the system can determine a significant deviation, thus identifying a conflict.

[0108] After identifying the conflict, the adjusted identification rules and thresholds are marked with conflict flags based on these results. These conflict flags can be Boolean symbols, confidence scores, or specific conflict type labels, used to clearly indicate the identification rules and thresholds used in this event identification. They may not be semantically completely consistent with the macro knowledge graph or historical reasoning results, and there is a risk that they need to be treated with caution or further reviewed.

[0109] Finally, conflict markers and preliminary macroeconomic impact assessment results are added to the populated event summary. That is, the final event summary not only includes detailed semantic event information and its identification parameters (including adjusted identification rules and thresholds) at the micro level, but also the preliminary macroeconomic impact assessment results corresponding to the event, as well as the results of consistency verification with the macroeconomic knowledge system. In this way, when used as a data indexing unit, the event summary simultaneously carries three types of key information: "the facts of the event," "the judgment of the macroeconomic impact," and "the reliability of the identification parameters."

[0110] The solution proposed in this application addresses the problem that existing technologies only focus on the micro level and lack a macro perspective and consistency verification in event summaries by introducing macro-impact assessment and conflict identification mechanisms into the event summary generation chain.

[0111] In some preferred embodiments, suppose a vertical search engine detects an abnormal increase in microplastic concentration in a certain sea area. After the rules and thresholds for identifying the event are structured and encoded and populated into the event summary, the system retrieves a macro-impact assessment rule from a preset rule base based on the type of "abnormal increase in microplastic concentration event." This rule can be a model used to predict the impact of different levels of microplastic pollution on the growth rate and reproductive capacity of specific fish species (e.g., cod). For example, if the identified event indicates that the microplastic concentration in the sea area has reached a certain critical level and has persisted for a week, the macro-impact assessment rule may provide a preliminary macro-impact assessment result: predicting that the average growth rate of cod in the sea area will decrease by 15%, and the survival rate of juvenile fish in the next reproductive cycle will decrease by 10%.

[0112] The system then compares this preliminary macro-impact assessment with a pre-stored macro-knowledge graph. Historical data recorded in the macro-knowledge graph shows that in several similar microplastic pollution incidents in the past, cod growth rates typically decreased by 5–8%, and juvenile survival rates decreased by 3–5%. Since the preliminary assessment results (decreases of 15% and 10%) are significantly higher than the historical experience range, the system identifies a conflict. Based on this conflict identification result, the system marks the identification rules and thresholds used to identify this abnormal increase in microplastic concentration with a conflict flag, such as "high conflict risk" or "requires expert review."

[0113] Finally, the populated event summary, in addition to recording the occurrence time, duration, key indicator statistics, and confidence level of the abnormal increase in microplastic concentration, will also include a "high conflict risk" conflict marker and a preliminary macroeconomic impact assessment result predicting a 15% decrease in cod growth rate and a 10% reduction in juvenile survival rate. When a user queries "the impact of microplastic pollution on cod production," the vertical search engine can not only retrieve the abnormal increase in microplastic concentration but also simultaneously present the corresponding macroeconomic impact predictions and conflict risk warnings, providing users with more comprehensive and risk-informed decision-making support.

[0114] In another embodiment of this application, S2243-1 further includes: S2243-11: Obtain current geographical location information; S2243-12: Obtain current season information; S2243-13: Obtain current fisheries activity data; S2243-14: Based on the identified semantic event type, current geographical location information, current season information, and current fishery activity data, match macroeconomic impact assessment rules from a pre-defined rule base.

[0115] Specifically, obtaining current geographic location information refers to acquiring the precise or approximate geographic coordinates of the area where an event occurs or is affected through the Global Positioning System (GPS), base station positioning, user input, or other geographic information system interfaces. This information may include longitude, latitude, administrative divisions, and specific water area names, etc., and its purpose is to provide spatial context for macro-impact assessments, enabling subsequent matching macro-impact assessment rules to be adjusted for specific sea areas or aquaculture zones.

[0116] Obtaining current seasonal information can be understood as determining the season (e.g., spring, summer, autumn, winter) based on the timestamp of an event, or more precisely, the specific month or ten-day period. This information is used to reflect the seasonal patterns of fisheries production and ecological processes, such as fish migration periods, breeding seasons, or periods of high temperatures. Its purpose is to introduce a temporal context into macroeconomic impact assessments, thereby reflecting the different macroeconomic impacts that the same semantic event may have in different seasons.

[0117] In practical applications, obtaining current fisheries activity data refers to collecting information such as fishing vessel operations, catch volume, aquaculture activity status, and catch species at the time of an event or within the affected area. For example, indicators such as the intensity of current fisheries activities, main target fish species, and aquaculture density can be extracted from fisheries management systems, Automatic Identification System (AIS) data, fishermen's logs, or market transaction data. The purpose is to reflect the actual status of current fisheries production and provide economic and production activity-level evidence for macroeconomic impact assessment.

[0118] In some embodiments, after obtaining current geographic location information, current season information, and current fisheries activity data, the system uses the above information along with the identified semantic event types as query conditions and matches them against a preset macroeconomic impact assessment rule base. Each rule entry in the rule base includes an event type field, a geographic region field, a season field, and a fisheries activity pattern field, for example, "Event type = abnormally high microplastic concentration; Region = a certain aquaculture area in the South China Sea; Season = summer; Fisheries activity pattern = peak season for grouper farming".

[0119] During rule matching, priority can be given to matching rule entries that are completely consistent with the current context. When no rule can completely match, the closest rule can be selected by gradually relaxing the conditions or by similarity matching. For example, the geographical area can be expanded from a specific aquaculture area to a larger-scale adjacent sea area, or the season can be expanded from a specific month to a seasonal level. In this process, the similarity between the current context and the rule entry in terms of geographical location, season, and fishery activity pattern can be calculated, and each similarity can be weighted and summed. The rule with the highest score is selected as the macro-impact assessment rule for the current semantic event.

[0120] For example, in a scenario where an "abnormal increase in microplastic concentration" occurs in a specific aquaculture area in the South China Sea during the summer of a certain year, and this area is in the peak season for grouper farming, the system first locates the corresponding aquaculture area based on the event's latitude and longitude information. Then, it determines the season as summer based on the timestamp and identifies the current peak season for grouper farming from fisheries activity data. Subsequently, the system searches the rule base for rule entries with the event type "abnormal increase in microplastic concentration," the region as "a specific aquaculture area in the South China Sea or its superior region," the season as "summer," and the fisheries activity mode as "peak season for grouper farming." If a perfectly matching rule exists, such as "a macroscopic impact assessment rule for the impact of increased microplastic concentration in aquaculture areas in the South China Sea during summer on the growth and survival rate of grouper," then that rule is directly selected for the macroscopic impact assessment. This rule can clearly indicate that under the high temperatures of summer, microplastics have enhanced toxicity to grouper fry, potentially leading to a decrease in grouper production of approximately 20% and increasing the probability of certain diseases.

[0121] Furthermore, based on the identified semantic event type, current geographic location information, current season information, and current fisheries activity data, macroeconomic impact assessment rules are matched from a pre-defined rule base. The pre-defined rule base is a set of macroeconomic impact assessment rules constructed based on historical data, expert knowledge, scientific research, and policies and regulations. Each rule is typically associated with a specific semantic event type, geographic area, seasonal conditions, and fisheries activity pattern. The matching process can employ rule-based reasoning, machine learning classification, or similarity matching methods, using "semantic event type + geographic location + season + fisheries activity data" as joint conditions to retrieve the macroeconomic impact assessment rule that best fits the current context, thereby avoiding the use of inapplicable general rules.

[0122] The proposed scheme incorporates current geographic location information, current season information, and current fisheries activity data, enabling the process of obtaining macroeconomic impact assessment rules to fully consider the specific spatiotemporal context of the event and the actual situation of fisheries production.

[0123] In some preferred embodiments, a specific example is given below. Suppose that in the summer of a certain year, an "abnormal increase in microplastic concentration" event occurred in a specific aquaculture area in the South China Sea, and this area was in the peak season for grouper farming. Traditional methods may only rely on the semantic event type "abnormal increase in microplastic concentration event" to extract a general macro-impact assessment rule for "the impact of microplastic pollution on fish resources" from the rule base, which cannot reflect the specificity of this sea area, this season, and this aquaculture mode. The solution of this application will first obtain the geographical location information of the event (a specific aquaculture area in the South China Sea), seasonal information (summer), and current fishery activity data (peak grouper farming season, farming density level, main farming stage, etc.), and then submit the combined conditions of "semantic event type: abnormal increase in microplastic concentration event + geographical location: aquaculture area in the South China Sea + season: summer + fishery activity: peak grouper farming season" to the preset rule base for matching. The system may match a specific macroeconomic impact assessment rule for "the impact of increased microplastic concentrations in the South China Sea aquaculture area during summer on the growth and survival rate of grouper." This rule can further explain the specific impacts of microplastics on grouper fry toxicity, feed utilization, and disease resistance under high water temperatures in summer, and provide a quantitative assessment of aquaculture yield and economic benefits. For example, the rule may predict that under the current concentration level and aquaculture conditions, grouper yield may decrease by 20%, and the morbidity rate of juvenile fish may increase significantly. Therefore, the obtained macroeconomic impact assessment rule is not only more scenario-specific but also provides precise and interpretable guidance for subsequent macroeconomic impact assessments, rather than simply giving generalized conclusions such as "environmental pollution may affect yield."

[0124] Reference Figure 2In this regard, a specific implementation of this application also discloses a fisheries data intelligent retrieval and indexing system, including: The acquisition and calibration module 1 is used to acquire multi-source micro-ecological data and perform collaborative calibration on the multi-source micro-ecological data. It identifies and corrects abnormal data based on the correlation between each data source and the preset normal fluctuation range to obtain calibrated micro-ecological data. The identification module 2 is used to identify semantic events based on the calibrated micro-ecological data and generate corresponding event summaries. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics and confidence information. The reasoning module 3 is used to update the semantic association knowledge representation structure based on the event summary. In the semantic association knowledge representation structure, it creates or updates nodes and relation edges that represent the first relationship between semantic events and entities related to the macro-output of fisheries. Based on the updated semantic association knowledge representation structure, it performs causal association reasoning on user queries about the impact of micro-environmental changes on the macro-output of fisheries. It also serves as the data index basis for the vertical search engine to retrieve and respond to queries. The first relationship is used to represent the influence relationship between related entities and their semantic events based on the time sequence.

[0125] This system, through its modular design, enables the collection, calibration, semantic event recognition, event summary generation, and updating of semantic association knowledge representation structures and causal reasoning of multi-source micro-ecological data. Specifically, the collection and calibration module 1 is responsible for data preprocessing, ensuring data accuracy and consistency; the recognition module 2 transforms the calibrated micro-data into event summaries with clear semantics; and the reasoning module 3 uses these event summaries to dynamically construct and update the knowledge graph and perform deep causal reasoning, thereby providing accurate and efficient data indexing services for vertical search engines. This system architecture effectively addresses the growing demand for refined management and forecasting in the fisheries sector, overcoming the challenges of existing technologies in micro-data integration and knowledge modeling.

[0126] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent retrieval and indexing of fishery data, characterized in that, include: Multi-source micro-ecological data is collected and the multi-source micro-ecological data is collaboratively calibrated. Abnormal data is identified and corrected based on the correlation between each data source and the preset normal fluctuation range to obtain calibrated micro-ecological data. Based on the calibrated micro-ecological data, semantic events are identified and corresponding event summaries are generated. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics, and confidence information. The semantic association knowledge representation structure is updated based on the event summary. Nodes and relational edges representing the semantic event and its first relationship with entities related to fishery macro-output are created or updated within the semantic association knowledge representation structure. Based on the updated semantic association knowledge representation structure, causal association reasoning is performed on user queries regarding the impact of micro-environmental changes on fishery macro-output. This serves as the data index basis for the vertical search engine to retrieve and respond to queries. The first relationship represents the influence relationship between related entities and their semantic events based on temporal sequence.

2. The intelligent retrieval and indexing method for fishery data according to claim 1, characterized in that, Before performing collaborative calibration on the multi-source micro-ecological data, the following steps are also included: Acquiring underwater acoustic signals; Extract acoustic features from the underwater acoustic signals; Identify the type of systematic environmental interference based on the acoustic characteristics; The multi-source micro-ecological data are compensated and calibrated according to the type of systematic environmental disturbance to obtain compensated and calibrated multi-source micro-ecological data; The multi-source micro-ecological data after compensation and calibration are collaboratively verified.

3. The intelligent retrieval and indexing method for fishery data according to claim 1, characterized in that, Based on the calibrated micro-ecological data, semantic events are identified, and corresponding event summaries are generated, including: Acquire multispectral fluorescence data; The microalgal bloom intensity index was calculated based on the multispectral fluorescence data. The optical interference compensation amount and ecological risk adjustment factor are determined based on the microalgal bloom intensity index, and the benchmark event triggering threshold used to identify abnormal increases in microplastic concentration are dynamically adjusted to obtain the dynamic event triggering threshold. Based on the dynamic event triggering threshold, an abnormal increase in microplastic concentration is identified from the calibrated microecological data. The abnormal increase in microplastic concentration is a type of semantic event. The dynamic event triggering threshold, the microalgal bloom intensity index, and the optical interference compensation amount are recorded in the corresponding event summary.

4. The intelligent retrieval and indexing method for fishery data according to claim 1, characterized in that, Based on the calibrated micro-ecological data, semantic events are identified, and corresponding event summaries are generated, including: Collect data on fishery activities; Obtain historical fisheries production data and records of historical ecological events; Based on the fishery activity data, the historical fishery production data, and the historical ecological event records, the identification rules and thresholds for semantic events are dynamically adjusted. Based on the adjusted recognition rules and thresholds, semantic events are identified from the calibrated micro-ecological data; Generate a corresponding event summary based on the identified semantic events, and record the adjusted identification rules and thresholds in the event summary.

5. The intelligent retrieval and indexing method for fishery data according to claim 4, characterized in that, Generate a corresponding event summary based on the identified semantic events, and record the adjusted identification rules and thresholds in the event summary, including: The adjusted identification rules and thresholds are structured and encoded, mapping the parameters of rule type, adjustment range and applicable conditions in the identification rules and thresholds to predefined short codes or enumeration values; A preset summary template is selected based on the type and importance of the identified semantic event. The summary template includes fields for representing the core information of the event and specific fields for carrying the structured encoding recognition rules and thresholds. The structured encoded recognition rules and thresholds are then filled into the selected summary template to obtain the filled event summary; The padded event digest is compressed to generate a compact event digest for transmission.

6. The intelligent retrieval and indexing method for fishery data according to claim 5, characterized in that, The adjusted recognition rules and thresholds are structured and encoded, including: Construct a parameter association graph based on the semantic relationships between the parameters in the adjusted recognition rules and thresholds; Identify the core parameter clusters and associated paths from the parameter association graph; The core parameter cluster and associated paths are mapped to predefined combined short codes or hierarchical enumeration values ​​to obtain the encoding results; The encoding result is verified according to the preset semantic consistency judgment rules to ensure that the encoding result is semantically consistent with the core parameter cluster and associated path.

7. The intelligent retrieval and indexing method for fishery data according to claim 6, characterized in that, The encoding result is verified according to preset semantic consistency judgment rules, including: Periodically obtain the latest fisheries policies, marine environmental early warning information, and historical event patterns; The semantic consistency judgment rules are dynamically adjusted based on the latest fisheries policies, marine environmental early warning information, and historical event patterns. The encoding result is verified according to the dynamically adjusted semantic consistency judgment rules.

8. The intelligent retrieval and indexing method for fishery data according to claim 5, characterized in that, The structured encoded recognition rules and thresholds are then filled into the selected summary template to obtain the filled event summary, which includes: Based on the identified semantic event type, obtain the corresponding preset macroeconomic impact assessment rules; The macroeconomic impact of the identified semantic events is assessed based on the adjusted identification rules and thresholds, combined with the macroeconomic impact assessment rules, to obtain preliminary macroeconomic impact assessment results. The preliminary macroeconomic impact assessment results are compared with the pre-acquired macroeconomic knowledge graph or historical reasoning results to identify whether there are any conflicts, and the conflict identification results are obtained. Based on the conflict identification results, conflict markers are applied to the adjusted identification rules and thresholds. The conflict marker and the preliminary macroeconomic impact assessment results are appended to the completed event summary.

9. The intelligent retrieval and indexing method for fishery data according to claim 8, characterized in that, Based on the identified semantic event type, corresponding preset macroeconomic impact assessment rules are obtained, including: Get current geographic location information; Get information about the current season; Obtain current fisheries activity data; Based on the identified semantic event type, the current geographical location information, the current season information, and the current fishery activity data, macroeconomic impact assessment rules are matched from a preset rule base.

10. A fisheries data intelligent retrieval and indexing system, characterized in that, include: The acquisition and calibration module is used to acquire multi-source micro-ecological data and perform collaborative calibration on the multi-source micro-ecological data. It identifies and corrects abnormal data based on the correlation between each data source and the preset normal fluctuation range to obtain calibrated micro-ecological data. The identification module is used to identify semantic events based on the calibrated micro-ecological data and generate corresponding event summaries. The event summaries are structured data containing event type, occurrence time, duration, key indicator statistics, and confidence information. The reasoning module is used to update the semantic association knowledge representation structure according to the event summary, create or update nodes and relationship edges representing the semantic event and its first relationship with entities related to the macro-output of fisheries in the semantic association knowledge representation structure, and perform causal association reasoning on user queries about the impact of micro-environmental changes on macro-output of fisheries based on the updated semantic association knowledge representation structure, and use it as the data index basis of the vertical search engine for retrieving and responding to the query. The first relationship is used to represent the influence relationship between related entities and their semantic events on a temporal basis.

Images