448 results about "Structure generation" patented technology

The invention discloses a report verification batch parallel execution optimization method based on a pluggable rule engine framework, and relates to the technical field of report verification, and the method comprises the following steps: receiving compliance rules, converting the compliance rules into a uniform data structure model, analyzing to generate metadata indexes, and supporting business rule set configuration; loading the matched business rule set based on the meta-information of the to-be-verified report, and registering the matched business rule set to an execution container; s3, receiving a to-be-verified report, analyzing a report structure, and generating a report field index table; performing field dependency analysis based on the report field index and the loaded business rule set, and constructing a rule field dependency graph; graph partition optimization and topological sorting are executed according to the dependency graph, and rule batches capable of being executed in parallel are generated; and uniformly collecting and outputting all rule execution results, hit states, abnormal records and performance indexes, and generating a structured verification report and an audit chain log. The financial statement compliance check execution efficiency can be improved.

The invention discloses a data management method based on a distributed storage system, relates to the technical field of data management, and is used for solving the problem of storage management behavior strategy mismatch. On the basis of dynamic perception of data access behaviors in a distributed storage system, a behavior feature vector fusing an access mutation rate, periodicity and an access span is constructed, behavior pattern recognition and strategy structure generation are completed, optimal deployment and hierarchical storage of copies are achieved through node resource state perception and construction of a cost function, and the strategy structure generation efficiency is improved. And after compression and consistency configuration are executed and the strategy falls to the ground, the compressibility and the offset trend of the behavior path are analyzed, scheduling management information is extracted, and a stable execution or strategy adjustment signal is generated, so that the problem of strategy execution mismatching caused by behavior perception deficiency in the distributed storage system is reduced, and the strategy execution efficiency is improved. The strategy closed-loop control and the resource scheduling optimization under behavior driving are realized, so that the data management efficiency and the strategy adaptation capability of the distributed storage system are improved.

The invention discloses a multi-modal-based software architecture intelligent design and optimization system, which relates to the field of intelligent design and optimization, and comprises the steps of performing modal perception and preprocessing on architecture design demand information input by a user, converting the architecture design demand information into structured semantic data, and optimizing and enhancing semantic expression by introducing a reinforcement learning strategy, forming a structured user intention vector set; through a semantic mapping and reasoning processing unit, user intention vectors in the user intention vector set are constructed into a semantic-component alignment graph, and graph structure modeling and confrontation generation algorithm combination are adopted to generate a candidate structure graph set; a structure diagram generation step in the structure generation module effectively improves the rationality, diversity and adaptation capability of an automatically generated structure, and provides core support for realizing automatic construction of a target-demand-oriented architecture structure.

The efficient generative task reasoning acceleration method based on the hybrid expert network comprises the steps that context information of related generative tasks is obtained, context data is preprocessed, and it is ensured that the data is suitable for model input; constructing an efficient reasoning model based on the hybrid expert network, and determining a reasoning process of the efficient reasoning model based on the hybrid expert network; training an efficient reasoning model of the hybrid expert network, optimizing model parameters, and storing an optimal model structure; and generating a reasoning result by using the optimal model structure. In a model improved by the method, an expert network is composed of a plurality of independent experts, and each expert is responsible for processing different input characteristics. The gating network dynamically selects which experts participate in the calculation according to input characteristics, and assigns an expert to each input by calculating a probability distribution. In this way, the gating network and the expert network are closely matched, it is ensured that only the most suitable expert is used, and therefore the calculation efficiency and the model performance are improved.

The invention discloses a 3D intelligent simulation generation method for clothing styles based on a model library, and relates to the technical field of digital twinning, and the method comprises the steps: carrying out the entity relation extraction and semantic mapping of multi-source clothing data, constructing a cross-modal knowledge graph, and extracting a physical constraint rule of the cross-modal knowledge graph; performing association reasoning and cross validation on the physical constraint rule and the version library to obtain a basic version template set; inputting the basic model template set into an intelligent clothing model, performing structured feature decoupling by the parameter analysis layer, dynamically adjusting a topological structure by the geometric generation layer, and generating 3D simulated clothing; and carrying out digital modeling on the basic model template set through a digital twinning algorithm, constructing a digital twinning body, carrying out physical simulation and optimization verification on the 3D simulation clothing by using the digital twinning body, and outputting a clothing style 3D simulation scheme. According to the invention, by constructing the pattern library and the intelligent garment model, the efficiency, accuracy and individuation level of garment 3D simulation generation are improved.

The invention relates to the technical field of three-dimensional model reconstruction, in particular to an existing building three-dimensional model reconstruction system based on point cloud data, and the system comprises a point cloud data processing module which carries out the octree recursive space subdivision of input point cloud data, builds a hierarchical index structure, and generates a multi-view fusion point set. In the method, octree recursive space subdivision is performed on the input point cloud data to establish a hierarchical index structure, so that the key area of the building surface can be positioned more accurately in subsequent data processing, data redundancy is reduced, and the precision and efficiency of three-dimensional model construction are improved; then voxel depth field data is obtained by calculating a truncation symbol distance function value of voxels, the description accuracy of a building surface structure is improved, and an initial surface grid is formed through contour surface extraction; meanwhile, plane slicing is conducted on the multi-view fusion point set in the preset building main axis direction, and contour details of all building structure sections are captured.

The invention relates to the field of computer systems, in particular to an operation stability testing method and system oriented to a heterogeneous computing system, and the method comprises the steps: constructing a system overall operation logic model comprising a resource topological structure, a task scheduling rule and a communication link matrix; identifying a key communication path according to the communication link matrix, analyzing a resource competition hotspot in combination with a task scheduling rule, and establishing an interference model library; calling a specific model instance in the interference model library according to a test target, configuring injection parameters, and dynamically applying interference in a system operation process; key performance indexes in the system operation process are collected in real time; based on the real-time monitoring data, a multi-index weighted scoring algorithm is adopted to calculate an overall stability coefficient, a robustness grade and a self-healing capability index of the system, and a stability evaluation result is generated; and generating a visual report according to the topological structure of the operation logic model, and supporting comparison and analysis of historical versions. Unified centralized management is realized, and test efficiency and evaluation accuracy are improved.

The invention discloses a large-scale heterogeneous computing power cluster training and pushing acceleration method, which comprises the following steps: S1, collecting operation characteristics of each node of a heterogeneous cluster, and constructing an equipment state parameter set; s2, loading a training and reasoning model, constructing an intermediate representation graph structure, and generating graph structure information; s3, inputting the graph structure information and the equipment state parameter set into a scheduling strategy model constructed based on a neural network ordinary differential equation, and generating an optimization decision result; s4, reconstructing a model graph according to an optimization decision result, and dividing the model graph into a training sub-graph and a reasoning sub-graph; s5, mapping the sub-graphs to heterogeneous nodes, performing resource matching and task distribution according to the equipment state parameter set, and generating a scheduling result; s6, collecting performance feedback information; and S7, inputting feedback information into the scheduling model to update parameters, and adjusting evolution function strategy parameters. According to the method, the scheduling acceleration of the training and pushing task of the deep model in the heterogeneous cluster is realized.

The invention discloses an HTML (Hypertext Markup Language) information extraction method and device based on a multi-LoRA cascade strategy, equipment and a medium, and relates to the technical field of HTML information extraction. The information extraction method comprises the following steps: acquiring a document and inputting a large language model; and the large language model judges whether a table is included. And if the table is included, calling a table processing LoRA adapter to extract table content and convert the table content into pseudo natural language description, calling a text processing logic module to extract adjacent text contexts of the table, and then performing semantic integration to obtain first text information. And if the table is not included, calling a text processing module of the table processing LoRA adapter to extract the text content, and obtaining the first text information. And calling a key information extraction LoRA adapter, and extracting the structured key value pair from the first text information to generate JSON data. And calling a nested structured generation module to convert the JSON data into a target sequence in a multi-layer nested JSON format.

The invention discloses a financial knowledge graph construction method and system based on artificial intelligence, and relates to the field of artificial intelligence data processing. The method comprises the following steps: performing multi-dimensional semantic analysis on a heterogeneous financial data source, and extracting a structured semantic fragment; constructing a financial entity perception unit, identifying a multi-granularity entity and generating a unique code; generating a preliminary relation graph based on the event cascade relation and the attachment structure, and injecting a semantic translation label; normalizing the atlas relationship through semantic separation and a label reconstruction mechanism to form a financial relationship network with consistent semantics; executing evolution increment iteration in combination with the newly added corpus, and dynamically updating nodes and edge sets; and performing semantic consistency and structural integrity evaluation on an iteration result, and outputting a stable financial knowledge graph structural body. By introducing a multi-factor semantic analysis model, a causal relationship modeling mechanism and a graph evolution iteration strategy, systematic improvement of the financial knowledge graph in the aspects of structural expression precision, semantic reasoning ability and dynamic adaptability is achieved.

The invention discloses a standard document writing method and system based on knowledge retrieval enhancement, and the method comprises the steps: carrying out the knowledge retrieval to obtain domain multi-source heterogeneous knowledge, constructing a domain knowledge graph and a content increment model, carrying out the semantic extraction, structure division, content segmentation and text granularity perception, determining first matching information and second matching information, and writing a standard document according to the first matching information and the second matching information. Matching a standard paragraph structure and a standard statement structure to generate a standard template, performing compliance examination to obtain a compliance text and a recommended text, performing exception screening and exception correction to obtain a compliance numerical value, inputting the compliance text, the recommended text and the compliance numerical value into the content increment model, and performing standardized conversion to obtain a standard material, and inputting the standard material of each section of statement into the standard template to obtain a standard document. According to the method, the standard document writing efficiency and accuracy can be improved, meanwhile, the method has good interpretability, and the method can be directly applied to a standard document writing system based on knowledge retrieval enhancement.

The invention belongs to the technical field of composite material design, and discloses a polymer composite material design method based on LLM knowledge database construction and reasoning optimization, and the method comprises the following steps: S1, target performance setting; s2, constructing a knowledge database; s3, constructing a reverse design model; s4, constructing a performance verification module; and S5, performing multi-round test-verification design loop optimization. According to the method, an intelligent design method integrating target setting, semantic knowledge database construction, a design assistant, a performance verifier, a prompt engineering module and a language model material knowledge system is provided, closed-loop reasoning and feedback of structure-mechanism-performance are achieved through multiple rounds of DA-VM collaborative optimization, and the design efficiency is improved. The method has the advantages of high design efficiency, strong structure generation rationality, accurate performance prediction, knowledge evolvable and the like, and is suitable for intelligent rapid development of the high thermal conductivity-wave absorbing-shielding-mechanical integrated functional polymer composite material.

The invention provides an unbiased missing modal learning method based on a multi-stage double diffusion network, and belongs to the field of computer vision, natural language processing and multi-modal information fusion. The method comprises the following steps: a multi-modal feature extraction module maps text modal data, image modal data and audio modal data into a unified potential representation space; the potential space representation of the three modals is used as the original feature of the missing modality and the original feature of the available modality; based on the missing modal original features and the available modal original features, training a multi-stage double-diffusion module through forward diffusion and reverse diffusion to obtain a trained multi-stage double-diffusion module, and based on the trained multi-stage double-diffusion module, performing global structure generation and modal conversion through available modal data to obtain a multi-stage double-diffusion model; and performing local detail optimization reasoning to generate missing modal data. According to the method, the problem of modal generation deviation in the existing multi-modal learning is solved, and the effectiveness of the multi-modal learning effect in the missing scene is enhanced.

The invention relates to a training method for providing an underwater vehicle main body structure design model and an underwater vehicle main body structure generation method. The training method of the underwater vehicle main body structure design model comprises the following steps: firstly, generating a noise adding main body structure based on a randomly initialized ordinal number stamp k, a sample main body structure and randomly initialized Gaussian noise generation data; secondly, determining predicted Gaussian noise data based on the ordinal number stamp k, the performance parameters of the sample main body structure and the noise adding main body structure by adopting a main body structure design model; and finally, determining a training loss function based on the generated Gaussian noise data and the predicted Gaussian noise data, and training the main structure design model based on the training loss function. Under the condition that only few sample main body structures exist, enough training sample data used for main body structure design model training can be generated, and model training under a small amount of sample main body structure data is achieved.

The invention relates to the technical field of data sampling, in particular to a neural network training data set generation method and device, equipment and a storage medium, and the method comprises the steps: obtaining the shape data, material data and target load position space of a part virtual model; determining boundary constraint conditions of the part virtual model according to the shape data and the material data; carrying out load point stratified sampling in the target load position space to obtain a plurality of sampling points and sampling parameters; according to the boundary constraint condition and the sampling parameters of the plurality of sampling points, constructing a simulation working condition, performing topological structure simulation under the simulation working condition, and determining a part topological structure corresponding to the simulation working condition according to a simulation result; and generating a training data set based on the simulated working condition and the corresponding model topological structure, training a neural network model by using the training data set, and generating a part topological structure under the target real working condition by using the trained neural network model. Therefore, the problem that the quality of the training data set is poor in the related technology is solved.

The invention provides a model code generation method and system based on a computational graph, and relates to the technical field of model code generation. The model code generation method comprises the following steps: in response to a user creation instruction, creating a target calculation graph based on a visual interface; analyzing calculation graph information of the target calculation graph; mapping a graph structure of the target calculation graph to a target structure of a target model based on the calculation graph information; and generating a corresponding model code based on the target structure of the target model. According to the method and the device, interaction with the user can be realized based on the visual interface, so that visual editing of the computational graph can be realized. The model code of the target model can be accurately generated according to the computational graph created by the user. According to the method, a close relation can be established through visual editing and code generation, so that the efficiency and accuracy of model development can be improved, the development cost can be reduced, and the flexibility of code generation and the adaptability to a complex model structure can be enhanced.

The invention discloses a big data comprehensive supervision platform based on intelligent decision, and the platform comprises a heterogeneous graph construction module which is used for constructing a supervision graph structure; the node feature generation module is used for generating node embedding representation; the structure aggregation module is used for generating a pooling graph on the basis of a DiffPool structure controlled by Transform; the graph enhancement module is used for generating a graph enhancement sample pair; the comparison training module is used for optimizing embedded representation; the anomaly detection module is used for identifying an abnormal node set; the risk map generation module is used for constructing a risk map structure; and the response scheduling module is used for generating and executing a supervision response instruction. According to the invention, the intelligent level and the structure expression ability of the supervision task are improved.

The invention relates to the technical field of bridge parameterization design, in particular to an interactive generation type bridge parameter design method and system.The method comprises the following steps that based on span, beam height and support counter-force parameters, the effect of the span on the beam height and the influence of the beam height on support counter-force are analyzed, a node and edge weight set is constructed through the parameter relation, a network structure is formed, and the network structure is constructed; and generating a parameter association network weight matrix. According to the method, by constructing the parameter association network, the visualization and analysis efficiency of the parameter relationship among the span, the beam height and the support reaction force is effectively improved, the design flexibility is improved through optimization and integration of the parameter constraint interval, and generation of the dynamic path has higher logicality and real-time response capability; by means of load distribution optimization, accurate adjustment of design parameters under the complex load working condition is ensured, the design robustness and global consistency are enhanced, the adaptability and reliability of parameter design are improved, and the efficiency and precision of bridge design are remarkably improved.

The invention relates to the technical field of gas pipeline construction, and discloses a polyethylene buried gas pipeline tracer line construction quality informatization management system. The system comprises a trace line feature extraction module, a topological graph construction module, a resource conflict analysis module, a quality node judgment module and a quality structure generation module. Wherein the feature extraction module obtains a detection signal and constructs an original feature sequence; the topological graph construction module establishes a feature point connection path based on the feature sequence and generates a feature topological graph; the resource conflict analysis module extracts resource allocation paths and screens conflict path combinations; the quality node judgment module collects quality evaluation labels and generates a construction quality label set; and the quality structure generation module counts the quality nodes associated with the labels, establishes an affiliation mapping structure and outputs a classification structure table. According to the system, informatization management of the construction quality of the tracer line is realized, and systematic control of the construction quality is assisted.

The invention relates to the technical field of green intelligent manufacturing, in particular to an AI-driven solid waste material and renewable energy collaborative utilization method. The method comprises the following steps of obtaining solid waste parameter data, and performing matching according to the solid waste parameter data and a preset material pedigree library to obtain solid waste material data; performing target confrontation generation according to the solid waste material data to obtain solid waste printable structure data; carrying out micro-scale process section analysis and macro-scale energy section analysis on the solid waste printable structure data to respectively obtain process section load data and energy section load data; and according to the process section load data and the energy section load data, printing parameters are generated, printing parameter data are obtained, and 3D printing operation is carried out. Full-link optimization from solid waste material parameter identification to structure generation and energy efficiency matching is achieved, and solid waste resources are promoted to be manufactured and converted into low-carbon, high-value and high-adaptation.

The invention provides a training method and system of a network community dynamic graph generation model and a graph generation method, and is applied to the technical field of data mining. The method comprises the following steps: acquiring a historical dynamic time sequence diagram of a target network, and carrying out step-by-step feature extraction on the historical dynamic time sequence diagram by utilizing a feature extraction module according to a time sequence to sequentially obtain a diagram embedding representation, a hidden variable and a hidden state; inputting the graph embedded representation into a community detection module to obtain a community detection result; inputting the hidden state into a community prediction module to obtain a community prediction result of the prediction time step; inputting the hidden variable and the hidden state of the previous time step into a topological structure generation module to obtain a generated network diagram; determining training loss based on the community detection result, the community prediction result of the prediction time step and the generated network diagram; and updating model parameters based on the training loss. According to the method, the problems that application scene requirements sensitive to dynamic changes cannot be met and the quality of generated dynamic graph data is not high in the prior art are solved.

The invention discloses a geometric model quadrilateral mesh generation method and system based on GPU parallelism, and belongs to the technical field of computer graphics. The method comprises the following steps: for a geometric model formed by a plurality of surfaces, obtaining parameterized coordinates of surface patches in parallel by a GPU thread, and performing spatial transformation on the parameterized coordinates by establishing a bounding box of the surface patches to obtain a surface patch splicing result; generating an original triangular mesh model based on a patch splicing result, and processing triangular units in the original triangular mesh model in parallel by the GPU thread to obtain a triangular mesh after cavity filtering and network geometric information of the triangular mesh; the GPU thread solves the network topology structure of the triangular mesh in parallel; the GPU thread performs parallel solution of the frame field based on the network geometric information and the network topology information of the triangular mesh; and generating a quadrilateral grid structure of the geometric model in combination with the frame field. According to the method, the grid structure generation efficiency is improved by improving the calculation performance of the curved surface frame field.

The invention discloses an improved particle filter map matching method based on rail transit 5G positioning, and the method mainly comprises the steps: 1) collecting original trajectory data obtained through 5G positioning, carrying out the sliding average processing according to a preset sliding window, generating a smooth new trajectory point sequence, and reducing the data redundancy; 2) constructing an improved particle filter map matching algorithm on the basis of the new track points, combining with subway tunnel fine map structure generation, considering space prior information such as a track center line and a track boundary, endowing particles with differentiated weights through a track center line judgment function and a track area judgment function, and determining the track area according to the differentiated weights; high-robustness matching of the track points and the tunnel environment is realized; and 3) in a particle filtering iteration process, setting an effective particle number threshold value, triggering a resampling operation when particles are seriously degraded, selecting a new particle set in a probability mode, avoiding falling into a degradation endless loop, and outputting a state posteriori estimation result in combination with a particle weight to generate an accurate trajectory.

The application discloses a standard document writing method and system based on knowledge retrieval enhancement, comprising the following steps: acquiring field multi-source heterogeneous knowledge by knowledge retrieval and constructing a field knowledge graph and a content increment model; performing semantic extraction, structure division, content segmentation and text granularity perception; determining first matching information and second matching information; matching standard paragraph structures and standard sentence structures to generate a standard template; performing compliance review to obtain a compliant text and recommended text; performing abnormality screening and abnormality correction to obtain a compliance value; inputting the compliant text, the recommended text and the compliance value into the content increment model to perform normalized conversion to obtain standard materials; and inputting the standard materials of each sentence into the standard template to obtain a standard document. The method can improve the efficiency and accuracy of standard document writing, has good interpretability, and can be directly applied to a standard document writing system based on knowledge retrieval enhancement.

The invention discloses a flexible business process automatic configuration method and system based on semantic intention recognition, and relates to the technical field of semantic recognition, and the method comprises the steps of semantic fragment extraction, process fragment atlas construction, process structure generation, process configuration visualization and flexible business process configuration. The method comprises the following steps: carrying out word segmentation, named entity recognition, semantic role labeling and relation extraction on natural language input by adopting a hierarchical guide type semantic decomposition combined extraction model, establishing a business semantic labeling system, and realizing modeling and extraction of semantic elements such as actions, objects, roles and conditions; on the basis, a process fragment graph improved by structural semantic fusion is constructed, and through node-level splitting, semantic tag binding and logic connection abstraction, process fragment graph data with tag attributes are generated, so that semantic understanding, process matching and cross-scene multiplexing capabilities are improved, and efficient, flexible and intelligent configuration of a business process is realized.

The invention discloses an automatic design and optimization system for a hydraulic speed control loop, which takes a large language model intelligent body as a core, realizes automatic design and optimization of the hydraulic speed control loop through semantic comprehension, knowledge retrieval and parameter optimization, and specifically comprises an LLM intelligent body, a data processing module and a data processing module, the analysis module is used for analyzing natural language input containing design requirements and performance targets of a user, generating a loop structure in combination with a local knowledge base and calling an optimization algorithm; the local knowledge base comprises a basic hydraulic loop, a key empirical relationship and a typical structure template which are artificially extracted; and the simulation and optimization module is used for executing a swarm intelligence optimization algorithm to optimize and adjust parameters of the loop structure. The invention constructs a hydraulic speed control loop automatic design and optimization system with semantic understanding, automatic generation and adaptive optimization capabilities, and aims to realize full-process intelligent design from user demand input to loop structure generation and performance optimization.

The invention discloses a server power supply system, a server, a topological structure generation method and a storage medium, and relates to the technical field of power electronics. According to the invention, a high-voltage centralized conversion-medium and low-voltage distributed distribution two-stage architecture and a physical isolation design of the first direct-current bus and the second direct-current bus are adopted, and an adaptive mounting platform is provided for load nodes such as a super-capacitor node and an energy storage node depending on the second direct-current bus; the mounting number of the load nodes can be determined based on the power required by the load nodes and the voltage and through-flow level of the second direct current bus, and high-voltage power supply conversion and power supply of the load nodes are separated, so that physical isolation of high-voltage direct current of the first direct current bus and low-voltage direct current of the second direct current bus is realized. The high-voltage direct current input is ensured to be safely, efficiently and stably distributed to each load node in the server after being converted. The problems of potential safety hazard and reliability risk caused by direct access of the high-voltage direct current to the load node, poor mounting adaptability of the load node and the like in the prior art are solved.

The invention discloses a modular robot automatic design method and system based on a large language model, and aims to realize full-process automation from natural language input to modular robot structure generation. The method comprises the following steps: constructing a module library comprising basic components and composite components; receiving a natural language description input by a user; performing semantic analysis by using a large language model, and automatically generating a component planning list; calculating spatial positions and connection relations of the components based on recursive assembly reasoning; and generating a three-dimensional assembly code and outputting the structure model. The system comprises an input module, a planner module, an assembler module, a module library and an output module, and all the modules cooperate to realize an automatic design process from a text to a structure model. According to the method, the use threshold of non-professional users is lowered, and the method has good adaptability and expansibility and is suitable for various robot structure design tasks.

The invention discloses a data warehouse table structure generation method and device, a medium and a product, and belongs to the technical field of data warehouses. According to the method, the Excel template configured with multiple pieces of source database table structure information is obtained, table field definition is automatically analyzed, the source data type is intelligently mapped and converted into the target data type according to the source database type, meanwhile, partition key fields are automatically recognized based on the preset strategy, and finally a complete and accurate data definition language (DDL) statement is generated. According to the method, remarkable technical breakthrough and value are achieved in practical application, the table building efficiency can be improved by 90% or above, batch processing of hundreds of tables is supported, and manual errors are reduced by 95% or above; mainstream database type conversion is comprehensively supported, and various data models such as Dores are adapted; the method has the capabilities of automatic type detection and conversion, intelligent key field selection and partitioning and bucket dividing strategy optimization.

The invention relates to the technical field of machine learning, and provides an emotion analysis method and device, computer equipment and a readable storage medium, and the method comprises the steps: carrying out the data preprocessing of a to-be-analyzed input text, and obtaining a word vector sequence; obtaining a metaphor feature library; according to the metaphor feature library, processing the word vector sequence by utilizing a long-short-term memory network based on an attention mechanism to obtain a metaphor recognition result; if the metaphor identification result shows that the metaphor exists, fusing the implicit meaning corresponding to the metaphor identification result into the semantic representation of the input text; and converting the processed input text into a target sequence according to the aspect term set so as to finish sentiment analysis on the input text according to the target sequence. Aiming at complex text analysis pain points in the fields of finance, medical health, old-age care and the like, an efficient and accurate solution is provided for deep understanding fed back by cross-industry customers through an innovative architecture of'metaphor recognition-semantic fusion-structured generation '.