41 results about "Sorting algorithm" patented technology

This invention discloses an online editing and parallel execution method and system for vehicle network test cases, relating to the field of vehicle network testing technology. It establishes an online editing environment via a web interface, enabling automatic generation of protocol code and providing testers with a convenient and efficient development experience. Regarding test case execution, it accurately analyzes the dependencies of test cases, uses a topology sorting algorithm to construct a reasonable execution plan, and adaptively switches between sequential and parallel execution modes. Combined with containerization isolation technology, it allocates an independent execution container for each test case, ensuring the isolation and stability of the testing environment. Simultaneously, it initializes a unified test execution context for each test case and establishes a communication connection with the vehicle network bus simulation engine through a standardized interface, achieving standardization and normalization of operations such as message sending and receiving, fault injection, and assertion verification. In variable management, it adopts a copy-on-write mechanism to achieve isolated sharing of project variables, avoiding data crosstalk.

The application provides a multi-agent dynamic workflow arrangement method and system based on an MCP protocol, and relates to the technical field of multi-agent collaborative work. The method comprises the following steps: constructing a standardized agent description and configuring an MCP protocol adapter; generating a task demand text through planning an agent, matching an optimal agent by using a cosine similarity algorithm, and obtaining an initial workflow arrangement graph including a task execution node and a task dependency relationship in combination with a logic analysis technology and a topological sorting algorithm; screening historical related content corresponding to the task execution node based on a context propagation strategy, determining a context dependency result according to the task dependency relationship, and generating context inheritance information by summarizing; accessing the optimal agent by using the MCP protocol adapter, and performing a task execution operation according to the initial workflow arrangement graph in combination with the context inheritance information to generate a task execution result, so that multi-agent efficient collaboration can be realized, and arrangement flexibility and interactive stability can be improved.

The application discloses an AI-driven media knowledge base retrieval method based on regional tagging, and particularly relates to the technical field of artificial intelligence, which utilizes a multi-modal large model to perform semantic segmentation on multi-modal media files in a knowledge base, identifies basic units carrying independent semantics, and generates a structured tag set, constructs a cross-modal semantic association network based on the generated structured tag set, wherein nodes represent different media regions, and edges represent semantic similarity, establishes cross-modal links by calculating the semantic similarity of regional tag vectors, vectorizes the current retrieval query of a user, combines historical queries and interaction behaviors to construct a user intent context, performs retrieval in the cross-modal semantic association network, obtains a first candidate set, optimizes the sorting result by using a collaborative reordering algorithm, shows the user retrieval results with high relevance, and dynamically adjusts weights and optimizes the semantic association network according to user interaction.

This invention discloses a method and system for generating integrated sales, delivery, and payment reconciliation reports in the industrial sector, relating to the fields of industrial MES / ERP report generation and financial reconciliation technology. This invention uses a single database stored procedure plus table-value parameters as a unified entry point, achieving full coverage of futures / spot contracts and self-owned / outsourced warehousing delivery scenarios. It achieves accurate alignment of payment and delivery records through a contract number partitioning and sorting algorithm, constructing a two-tiered automatic balance accounting system for contracts and customers, and automatically generating standardized reconciliation reports. This invention solves the problems of cumbersome manual operation, inaccurate data, and untimely report updates in traditional reconciliation methods, improving reconciliation efficiency by over 90% and achieving 100% data accuracy. It is compatible with existing ERP / MES systems in industrial enterprises, has low implementation costs, and is suitable for daily reconciliation, data statistics, and business analysis scenarios in the industrial sector. This invention has been fully implemented in actual industrial ERP systems and can operate stably in real production environments for a long time, possessing mature engineering value and industrial promotion capabilities.

The application discloses a signal reading method, system, device and medium for a multi-target array test strip, and relates to the technical field of fluorescent immunochromatographic detection. The method collects an RGB fluorescent image of the multi-target array test strip; a binary image is obtained through grayscale processing and adaptive threshold segmentation; an effective signal contour is extracted; a spatial position calibration and automatic sorting of discrete fluorescent signals are performed through an intelligent sorting algorithm combining DBSCAN clustering and principal component analysis; non-uniform background interference is removed through a robust local regression method to obtain a purified signal; and finally, real RG ratio signal values of each target point are extracted. The application solves the problems of difficulty in multi-target positioning and incomplete background removal of the existing array test strip signal reading method through an intelligent algorithm, realizes accurate positioning, automatic sorting and efficient purification of multi-target fluorescent signals, and significantly improves detection accuracy. The application is suitable for quantitative analysis of multi-target immunochromatographic test strips.

The application relates to the technical field of mechanical arm obstacle avoidance, and particularly discloses a mechanical arm obstacle avoidance method and system based on vision. The method comprises the following steps: collecting environmental vision data to identify auxiliary potential obstacles and construct an attribute matching list; combining task real-time execution stage features, improving an ideal solution sorting algorithm to construct an auxiliary adaptation degree evaluation model, determining an auxiliary target obstacle after sorting, designing an auxiliary scheme and extracting a path planning core constraint; positioning an auxiliary function node and an initial path, and outputting a cooperative motion trajectory through quadratic programming optimization; and calculating a failure judgment coefficient based on the auxiliary adaptation degree and a demand achievement rate, and performing local reconstruction or updating of the trajectory according to scenes. The system comprises an obstacle identification module, a path planning module, a trajectory adaptation module and a trajectory updating module. The application converts obstacles from obstacle avoidance objects into auxiliary resources, which is beneficial to improving the task efficiency of the mechanical arm.

The application discloses a kind of anti-collision control systems applied to medical transport vehicle with self-cleaning function, comprising: real-time environmental data is collected from the periphery of vehicle by sensor and clean state index is obtained from vehicle equipment, and comprehensive monitoring vector is obtained;Priority sorting algorithm is used to weight adjustment on leading task type, and high priority sub-vector is separated from comprehensive monitoring vector for subsequent processing;The first instruction is selected from the execution queue and simulated verification is carried out in combination with vehicle dynamic parameters, and the adjusted queue after verification is obtained;According to the updated comprehensive monitoring vector, the priority sorting algorithm is repeated and the final decision path is generated, whether the problem caused by task conflict is solved to output the balanced control signal, the application realizes the maximum cleaning task efficiency under the premise of ensuring driving safety, avoids the safety hazard or cleaning interruption caused by single task priority, and improves the intelligent adaptive ability and overall operation effect of vehicle in complex scene.

The application discloses a kind of graph node mixed traversal sorting method and system of workflow visualization, it is related to data visualization technical field.The method includes: obtaining workflow graph and initializing data structure;From start node, perform depth-first mixed traversal, repeat detection and block to node in the traversal process;After traversal is completed, the coordinates of each node are calculated, and the secondary sorting is executed to the node list, finally, continuous serial number is distributed and the sorting result is output.The application realizes zero repeat processing by throttling blocking mechanism, realizes zero data backflow by branch anti-backflow control mechanism, improves the sorting accuracy by secondary sorting algorithm, has excellent real-time response performance, and can be widely applied in workflow visualization, ETL task arrangement, low-code development platform and other fields.

The application discloses a multi-aircraft multi-site delivery scheduling planning method, a storage medium and equipment, and belongs to the technical field of aircraft delivery scheduling planning.The application solves the problem that the existing method cannot obtain an optimal scheduling scheme for a large-scale aircraft to visit multiple sites.The application considers a scheduling planning problem of multiple aircrafts successively arriving at multiple sites, considers processes such as unloading, loading, refueling and maintenance, optimizes the visiting sequence of the aircrafts to each site, establishes a multi-aircraft multi-site delivery scheduling model, proposes a discrete data iterative sorting algorithm based on a neural network for a large number of aircraft sorting complex problems, solves the sorting problem of high-dimensional input through layer-by-layer decomposition iteration, realizes the target of the shortest total time consumption of material delivery, and obtains a better delivery scheduling scheme.The method can be applied to multi-aircraft multi-site delivery scheduling planning.

This invention discloses a method for automatically generating reference schematics for electronic devices, an electronic device, and a storage medium. The method includes: extracting the set of actual chip models corresponding to the underlying physical point diagram of the electronic device; assigning uniformly standardized virtual standard reference designators to components in the chip model set based on a global standard mapping dictionary, and establishing a heterogeneous data mapping table; using the chip model set as the query key, initiating a batch search to a general circuit module library to retrieve matching general circuit data packages; dynamically replacing variable placeholders in the general circuit data packages with virtual standard reference designators based on the heterogeneous data mapping table; performing topological sorting on the replaced general circuit data packages according to classification weight values ​​using a topological sorting algorithm to generate a multi-page structured logic reference schematic set, and generating a dynamic directory with interactive jump anchors based on the multi-page structured logic reference schematic set. This method can solve the problem of low repair efficiency caused by the lack of original manufacturer circuit diagrams.

The application belongs to the technical field of strain field inversion of structural health monitoring, and provides a displacement reconstruction method introducing tensor component hybrid theory and a measuring point layout optimization method, comprising: first, establishing a four-node inverse shell element mapping function, deducing an element strain-displacement conversion relationship, then introducing a tensor component hybrid theory to construct a transverse shear strain function meeting a coordinated deformation, then based on a weighted least square criterion, using measured strain data to preliminarily reconstruct a displacement field, then expanding sparse measuring point strain data through a radial basis function to generate virtual strain data, optimizing a pseudo stiffness matrix and a pseudo load vector, and improving the displacement reconstruction accuracy, finally taking the displacement reconstruction error and the number of strain measuring points as fitness functions, using a non-dominated sorting algorithm to optimize the measuring point layout, and obtaining an optimal scheme considering the accuracy and the layout efficiency. The application can improve the displacement reconstruction accuracy under sparse measuring points, and considers the reconstruction accuracy and the layout efficiency.

This invention discloses a method for deep memory optimization in AI agents. The method includes: constructing a hierarchical memory storage architecture based on Markdown files, storing long-term memory and short-term memory divided by date; employing a hybrid retrieval strategy combining vector semantic search and BM25 keyword retrieval, and combining MMR deduplication and time-series decay strategies to obtain a candidate memory set; calculating a comprehensive score for the candidate memory set using a dynamic relevance ranking (DRS) algorithm that integrates five dimensions: semantic similarity, temporal proximateness, importance, access frequency, and task alignment; outputting the final retrieval results in descending order of the score; and adaptively updating the weight coefficients of each dimension based on retrieval feedback through an online learning algorithm; and achieving automatic triggering of memory capture and recall functions. This invention can improve the adaptive capability and task completion rate of AI agents, and effectively reduce system storage and operating costs.

The application discloses a method for improving simulation speed of physical quantity of cross-field diode, relates to simulation technology of vacuum electronic device, and is proposed in view of the long simulation time in the prior art. Before solving the electric field force corresponding to each electron particle, the numbering is sorted in sequence from the cathode to the anode according to the positions of all the electron particles at the current time, the numbering is corresponding to the positions of the electron particles, and then the electric field force corresponding to each electron particle is solved according to the partial solution of the diode potential. The method has the advantages that the solving step of the electric field force corresponding to the partial solution of the diode potential in the cross-field diode is improved, the fast sorting algorithm is combined, the simulation speed is greatly improved compared with the general meshless numerical algorithm, and the method is suitable for the vacuum space of the cross-field diode in which the electron particles are injected from the cathode to the space with an arbitrary initial speed and an arbitrary magnetic field value.

The purpose of this invention is to address the problem of weak implicit logic recognition in financial texts by providing a system that deeply integrates external knowledge retrieval with financial analysis logic. The system includes the following steps: Step S1: Obtain the financial news text to be analyzed; Step S2: Based on a preset hybrid retrieval strategy, retrieve supplementary information related to the financial news text from a dynamic knowledge base. The hybrid retrieval strategy includes semantic retrieval and keyword retrieval, and optimizes the retrieval results through ranking fusion and re-ranking algorithms. This invention, through the synergistic effect of retrieval enhancement generation and model fine-tuning, improves the accuracy of financial sentiment analysis from 77.53% to 83.27% on the Chinese benchmark set, and by up to 12.82 percentage points on the English benchmark. Through direct preference optimization training, the model's deep thinking ratio is improved by an average of 33.3%, enabling the model to spontaneously perform multi-dimensional, forward-looking professional financial reasoning, rather than simple keyword matching.

This invention belongs to the technical field of e-commerce logistics sorting equipment, specifically relating to an intelligent sorting production line structure for e-commerce logistics. It includes an inclined conveyor, a main conveyor track, and branch sorting tracks, as well as a multi-path sorting unit. The inclined conveyor is fixed at the inlet of the main conveyor track, and branch sorting tracks for sorting express packages are correspondingly arranged on both sides of the main conveyor track. A buffer transition section is provided on the inclined conveyor near the inlet of the main conveyor track, and an intelligent identification and positioning unit is correspondingly arranged on the buffer transition section. Through the dual identification of a 3D vision scanning component and an RFID reading and writing component, combined with a pressure sensor, comprehensive and accurate collection of goods information is achieved. Combined with intelligent sorting algorithms and dynamic path planning, the sorting cycle can reach over 1200 pieces / hour, with a sorting accuracy rate ≥99.9%, significantly better than traditional sorting equipment.

This invention discloses a rule-supervised image recognition verification learning method and apparatus, comprising: a symbol recognition stage, in which an image set is input into a neural network model to obtain a set of symbol probabilities predicted by the model, a score index for evaluating the quality of symbol combinations is determined, and a heap structure is established based on the score index according to a dynamic combination ranking algorithm; and a dynamic verification stage, in which a knowledge verification function V is used to verify the symbol combination. KB The heap structure is dynamically maintained by sequentially determining whether the top symbol combination satisfies the knowledge rule based on the order of symbol combinations in the heap structure, until the first verified combination is found, thus finding the combination that satisfies V. KB The optimal symbolic combination solution with "constraint and maximum score index" is found; feedback optimization stage: the model output distribution is adjusted according to the prior distribution information to alleviate the inference shortcut problem, and the found optimal combination solution is used as supervision information to feed back to the model for gradient optimization, thus completing the image recognition training with rules as supervision information.

A system and method of source and type-independent organization, presentation, and management of security detection artifacts, with heuristic and ML-based (machine learning) prioritization algorithms, providing for unified threat detection management and accurate prioritization without requiring an expert user. This is accomplished using an alternative presentation paradigm underpinned by a combination of heuristics and machine learning. The benefits include greater information gain; improved speed of decisioning cybersecurity detection data; elimination of the split-brain problem where different classes of detection artifacts are stored and managed in different systems; and reduced need for expert users. In cases where machine learning and / or anomaly detections are not available for clustering.

This application provides an optimization method and related equipment for new energy vehicle charging scenarios on highways, belonging to the field of strategy optimization technology. The method includes: acquiring static basic data and dynamic real-time data of highway service areas; generating multiple candidate charging guidance strategies in a discretized decision space based on the static basic data and dynamic real-time data; rapidly evaluating each candidate charging guidance strategy to obtain a three-dimensional objective vector corresponding to each strategy; the three-dimensional objective vector includes a power grid load stability objective, a user waiting time objective, and a traffic congestion objective; selecting the optimal solution set from all candidate charging guidance strategies based on a non-dominated sorting algorithm, and determining the final charging guidance strategy from the optimal solution set using the ideal point method; executing the final charging guidance strategy and performing rolling optimization based on real-time feedback data. This application embodiment can ensure the high efficiency of new energy vehicle charging guidance operation in highway service areas.

The application relates to a kind of measurement and adjustment method and system for large spherical shell flange assembly, and relates to a kind of measurement and adjustment method and system for spherical shell flange assembly.The present application is to solve the problems that flange cannot be aligned to theoretical pose caused by flange hole manufacturing error in traditional assembly technology, it is difficult to quickly evaluate the pose of flange in the process of assembly, flange positioning is difficult, assembly space is limited, leading to the difficulty of adjusting flange through mechanism.The steps of the method include evaluating the optimal reachable pose of flange based on the manufacturing error of flange mounting hole using improved non-dominated sorting algorithm;calibrating flange coordinate system;positioning based on the measured value of the distance between the corner point of flange mounting hole and the center of the sphere;establishing theoretical flange coordinate system in the center coordinate system of the sphere according to the assembly process requirements of flange;obtaining the theoretical coordinate information of auxiliary positioning point in the center coordinate system of the sphere.The application belongs to the field of high-precision assembly and measurement technology of flange parts.

The application discloses a kind of avionics system cross-domain collaborative efficiency real-time quantitative evaluation method and system, it is related to avionics simulation test technical field.The system constructs the distributed verification environment including space-time consistency synchronization hub, deterministic fiber reflection memory network and multiple heterogeneous subsystems.The method comprises: constructing evaluation model based on tactical scenario and configuring collaborative efficiency index system;Real-time acquisition of heterogeneous data marked with global absolute time stamp is carried out through deterministic network, wherein physical feedback time is obtained by monitoring physical signal jump of real device through FPGA;Space-time alignment is carried out on data using interpolation and sorting algorithm;Real-time quantitative values such as collaborative action synchronization error, task chain closed loop success rate and reconnaissance-dry-communication collaborative efficiency index are calculated;And dynamic closed loop feedback is carried out based on evaluation result.The application solves the problems that time sequence is difficult to align, efficiency is difficult to quantify and environment reconstruction is complex in heterogeneous avionics system collaborative test.

This invention discloses a multi-GPU bitone sorting Top-K optimization method based on shared memory and a dynamic threshold queue. The method includes the following steps: reconstructing the bitone sorting Top-K into a two-stage iterative execution logic; constructing a global dynamic elimination threshold queue, extracting candidate elimination thresholds from each data block and storing them in the queue; designing an adaptive update mechanism for the global elimination threshold queue, dynamically calculating the elimination threshold for the current round based on the distribution of candidate elimination thresholds in the queue; eliminating elements in the sorted data that do not meet the conditions according to the elimination threshold, and reorganizing the remaining data; when the data size shrinks to below a preset threshold, sending the remaining data back to the host to perform a complete bitone sort, outputting the final Top-K ordered sequence. The purpose of this invention is to address the challenges of computational power and memory bottlenecks on single-GPU platforms and difficulties in multi-GPU expansion when processing massive data sorting queries, by proposing a multi-GPU bitone sorting Top-K optimization method based on shared memory and a dynamic threshold queue. This method achieves rapid data elimination and dynamic recombination by reconstructing the bitonetic sorting algorithm and introducing a candidate threshold extraction mechanism and a global dynamic threshold queue in shared memory. This significantly improves the resource utilization and parallel efficiency of multi-GPU platforms, thereby accelerating the processing of large-scale Top-K query tasks.

The application discloses a flexible printed circuit board wiring optimization method based on laminar flow closing sequence, and is oriented to the same layer high-density parallel wiring scene after layering. The method constructs a wireable domain based on the plate shape boundary and the forbidden wiring area, extracts a skeleton channel network, and quantifies the available width on both sides of the channel to form a channel graph with capacity constraints. The same layer line group is expanded into a wiring instance according to the number of parallel roots, a wiring sorting algorithm based on the laminar flow closing sequence is proposed, and the wiring priority of the wiring instance is determined through the 'close first and open later' stack matching rule of the clockwise scanning of the port. Then, the incremental wiring is performed under the capacity constraint according to the sequence, a geometric path meeting the line width, line spacing, safety margin and minimum bending radius is generated, and space occupation update and conflict avoidance are realized through area write-back. The method can reduce the risk of shared channel crossing and crossing, improve the success rate of automatic wiring and manufacturability. At the same time, the manufacturability and regularity of the wiring result are optimized.

The application provides a news synthesis method and system based on traditional algorithms and large models, and relates to the technical field of natural language processing and text generation. The method comprises: obtaining news texts of different data sources and preprocessing; using an improved text segmentation and sorting algorithm, performing semantic synthesis on the preprocessed news texts, generating a comprehensive news summary text and performing semantic role labeling to obtain a structured fact anchor point; constructing a structured prompt word template and configuring a low randomness parameter, and calling a large-scale language model to generate a preliminary news draft that meets the fact boundary according to the fact anchor point; performing quality detection on the news preliminary draft to obtain a comprehensive score report and a quality defect label of the news preliminary draft; using a large-scale language model to generate structured feedback of the news preliminary draft, and iteratively optimizing the news preliminary draft according to the feedback result to generate a news manuscript. The application realizes automatic generation of news with high accuracy, high consistency and clear structure.

The application provides a knowledge graph-based alarm root cause positioning method and device, medium and equipment, relating to the technical field of big data. The method obtains a plurality of historical alarm data, constructs a network device topology graph based on the cleaned and labeled device identifiers and device link relationships, and the network device topology graph is used to represent the connection relationship between devices. Feature engineering is constructed and data is compressed, alarm type entity extraction and relationship extraction are performed, an alarm type causal knowledge graph is constructed, and the alarm type causal knowledge graph is used to represent the generation relationship between alarm types. According to the network device topology graph and the alarm type causal knowledge graph, an alarm propagation graph is generated based on a fusion rule. According to the alarm propagation graph, an alarm root cause propagation link graph is obtained by reasoning based on a graph search algorithm and a topological sorting algorithm, network device alarm root cause positioning is performed, and automatic positioning of the root cause is realized, which can improve the root cause positioning efficiency and reduce the labor cost.

This invention relates to a method for analyzing the behavior of stored grain pests based on YOLOv5 and an improved SORT algorithm, belonging to the technical field of stored grain pest behavior analysis methods. This method utilizes a camera to collect and filter videos of stored grain pests crawling, and processes the videos frame by frame. First, each frame is preprocessed using OpenCV to detect the outline of the stored grain pests and extract target coordinate data. When target overlap or loss occurs in a frame, resulting in detection failures for certain stored grain pests, the frame is segmented and fed into the YOLOv5 target detection network for re-detection, optimizing the detection results. This method solves the problems of low accuracy, susceptibility to environmental factors, and poor detection performance when overlap occurs in existing stored grain pest behavior analysis methods, making it particularly suitable for the needs of stored grain pest behavior analysis.

This invention discloses a method and system for playing back air traffic control data based on message timing. The method includes: a reading thread extracting the timestamp and hexadecimal message data of each line of the message based on a two-level timestamp extraction strategy; a decoding thread extracting the message and sorting it using a sliding time window double threshold sorting algorithm before sending it to the decoding queue; a sorting thread transmitting the message data; and a sending thread and a dynamic scheduling algorithm based on a reference time calculating the precise sending time of each message and sending it to the target system for playback. The advantages of this invention are as follows: (1) The playback timing accuracy is naturally matched with the protocol accuracy, and the time error is strictly controlled within 1 / 128 seconds; (2) It effectively solves the trajectory distortion caused by the disorder of the original data; (3) The system resource occupancy rate is low, supporting high-concurrency playback tasks; (4) The startup delay is short, meeting the fast response requirements of the playback system; (5) The processing throughput is high, adapting to high-density air traffic control data playback application scenarios.

This invention discloses an intelligent optimization method for the coordinated scheduling of cascade reservoir groups oriented towards multi-objective equilibrium, belonging to the field of swarm intelligence technology. The method includes: real-time acquisition of time-series monitoring data, simultaneously acquiring hydrological ensemble forecast data; generating a set of feasible scheduling trajectories satisfying rigid constraints using a fast non-dominated sorting algorithm, and generating an initial scheduling strategy population by combining a typical scenario library and a transfer learning mechanism; using a robust optimization algorithm to perform rolling solutions on the feasible scheduling trajectories and the initial strategy population, generating water level control intervals and unit output command sets; and using a multi-objective decision visualization platform, displaying the water level change intervals and risk probabilities of each scheme with a 3D interactive interface, generating Pareto front schemes for dispatchers to select and execute based on real-time preferences. This invention can effectively improve the scheduling robustness and multi-objective equilibrium capability of cascade reservoir groups under hydrological uncertainty, providing intelligent decision support for the safe and efficient operation of reservoir groups.

The application discloses a kind of based on minimum operating gene unit and digital twin dining operating real-time deduction and closed-loop control method and system. Including: the dining operating entity is disassembled into minimum operating gene unit represented with characteristic vector;Correlation constraint matrix is constructed, and its triplet definition space mutual exclusion time sequence dependence and resource binding relationship and priority;With gene unit as node, construct digital twin, and gene state is synchronized in real time through multi-source heterogeneous data stream;In response to trigger event, the multi-branch discrete event simulation of gene state combination snapshot is carried out, and gene state transition is simulated in parallel based on constraint matrix and priority sorting algorithm, and gene state evolution function is formalized description and coefficient is dynamically optimized by machine learning;According to multi-objective weighted evaluation function, select optimal deduction branch, convert into instruction set by mapping rule, and execute through API to physical terminal. Realize perception-decision-execution closed loop, decision second level, error <5%, utilization rate is improved by 15%-25%.

The application discloses a kind of knowledge graph construction methods based on fine-grained retrieval and reverse restoration self-error correction, specifically: first, standard reference case library is constructed, and standard reference vector is calculated.Then, the long text S to be measured is disassembled into sentences to be processed;Each sentence is vectorized, and the similarity score of its standard reference vector is calculated, and the first standard reference case of each sentence is screened;Through large language model, the non-standard logical relationship contained in each sentence is extracted and vectorized, the cosine similarity is calculated, and the candidate mapping is obtained by descending arrangement, after summarizing, double-feature reordering is carried out using multi-round greedy reordering algorithm, and the first standard reference case is screened out, and the single sentence is spliced into large language model, and triple extraction is carried out;Finally, the relationship triple set of all sentences constitutes knowledge graph.The application improves the precision and recall rate of triple extraction under the premise of avoiding redundancy.