139 results about "Semantic equivalence" patented technology

The invention discloses a multi-source legal knowledge graph construction and intelligent retrieval method for consumer right protection, and relates to the field of legal knowledge graphs and retrieval, the method comprises the following steps: constructing a multi-source legal knowledge graph, unifying semantic equivalence legal concept nodes through an entity alignment mechanism, and adding region and industry source labels to the nodes; analyzing the natural language problem of the user, and extracting related legal entities, dispute elements and region and industry context constraints; by adopting a hybrid retrieval strategy in which semantic vector retrieval and keyword retrieval are parallel, pre-filtering and limiting a retrieval range in combination with the context constraint, and generating a candidate legal clause set; executing multi-hop path reasoning on candidate legal terms based on the knowledge graph to generate an association path with node contribution weights; and outputting traceability answers with legal clause references and case sources. According to the method, the coverage, the accuracy and the credibility of the legal consultation service can be remarkably improved, and the method has relatively high application value.

The invention discloses an automatic SQL tuning method and system combined with a large language model and a database optimizer, computer equipment and a storage medium, and the method comprises the steps: generating a query fingerprint according to an SQL query statement input by a user, detecting whether there is an optimization result matched with the query fingerprint in a cache, and only when there is no optimization result, executing the step 1; if yes, the SQL query statements input by the user and the optimization instruction are input into the large language model, and a plurality of candidate SQL query statements are generated; performing semantic equivalence verification on the candidate SQL query statements and an SQL query statement input by a user, and screening out semantic equivalence candidate SQL query statements; submitting the screened candidate SQL query statements to a database optimizer, and obtaining an execution plan and cost information of each candidate SQL query statement; calculating an evaluation score of each candidate SQL query statement, and selecting the candidate SQL query statement with the highest evaluation score as an optimization result; and writing the optimization result and the related execution data into a cache, and returning the optimization result and the related execution data to the user.

The invention discloses a data integration and fusion method and system based on a large model, and the method comprises the steps: taking the metadata information of initial multi-source data and representative sample data as the input context of a pre-trained LLM large language model, and carrying out the training through a Prompt prompt project guide model, thereby obtaining a plurality of enhanced data source mode descriptions; inputting a plurality of enhanced data source mode descriptions into an LLM large-scale language model, enabling the model to be a data integration engine through a Prompt prompt project, guiding the model to recognize semantic equivalence fields among different data source modes, and outputting a dynamic mode mapping scheme; analyzing a conflict reason and reasoning a conflict resolution strategy, and outputting a data cleaning suggestion and a data analysis result; and loading the fused data into the target data storage. The method can adapt to complex and changeable data structures, and the flexibility and efficiency of data integration are improved.

The invention discloses a computer interactive programming platform automatic evaluation system based on deep learning, and particularly relates to the technical field of deep learning, the system comprises five modules, namely a structure diagram construction module, a semantic diagram generation module, a dynamic evaluation trigger module, a semantic solution identification module and an exponential deduction feedback module, and full-process evaluation from code structure abstraction to semantic modeling is realized. The system dynamically controls a deep evaluation process through a structure matching rate and a logic deviation degree, identifies a potential innovation solution, generates a semantic equivalence degree and a path innovation index, outputs a grade in combination with a neural network score, and embeds a high-innovation sample into a training buffer area to update a model, so as to realize intelligent evaluation and adaptive optimization; according to the method, the modeling precision of a program structure and a control flow is improved, dynamic regulation and control of semantic understanding and an evaluation process are realized, a potential innovation solution can be identified, and model self-updating is realized through scoring grade feedback, so that the intelligent analysis and adaptive evolution capability of an evaluation system on complex programming behaviors is enhanced.

The invention provides a data lake construction method and system based on multi-source distributed data, and relates to the field of data lake construction. The method comprises the following steps: firstly, deploying a data acquisition agent at each node of a distributed data source and establishing a distributed collaborative mechanism, and performing structured analysis on data through the data acquisition agent to obtain data characteristics and establish a relationship between fields; and then semantic description information is constructed, and an RDF triple is generated based on the information and stored in a source end node. And then inputting the RDF triple into an ontology registration service for processing to generate a distributed ontology, and analyzing a corresponding relationship between the RDF triple and concepts in the distributed ontology to obtain a semantic equivalence rule and a mapping rule. And finally, organizing the distributed data sources into a unified data view according to the rules, establishing a distributed index, configuring an access permission, and finally constructing a data lake containing the unified data view, the distributed index and the access permission. According to the method, the performance bottleneck problem caused by centralized processing is avoided.

The invention discloses a cross-language code semantic alignment method, which constructs a shared semantic space through a unified abstract syntax tree (AST) and a graph matching network (GMN) so as to reduce the difference of different programming languages in syntax structure and node representation. The method comprises the following steps: (1) mapping a multi-language AST node to a unified general label set and performing structure enhancement; (2) performing node feature coding on the unified AST, and realizing cross-language interaction in combination with a cross-graph attention mechanism; (3) node representation is generated through intra-graph loop updating, and an overall semantic vector is obtained through global attention pooling; and (4) through comparative learning training in the shared space, the distance between semantically equivalent positive sample vectors is shortened, and the distance between non-equivalent negative sample vectors is shortened, so that the discrimination capability of cross-language semantic representation is enhanced. According to the method, the semantic consistency of the functional level can be effectively captured, and the accuracy and efficiency of cross-language code understanding, multiplexing and retrieval are remarkably improved.

The invention discloses a communication data intelligent retrieval method and platform based on semantic analysis, and relates to the technical field of natural language processing, and the method comprises the following steps: after receiving an original query text of a user, binding an association retrieval permission strategy based on a user ID, loading a user portrait to complement semantics to obtain a plurality of semantic equivalence complement statements, and storing the semantic equivalence complement statements in a database; the method comprises the following steps: extracting feature words of communication data, constructing a plurality of hierarchical retrieval feature trees in combination with a domain knowledge graph, then matching a distributed retrieval unit, carrying out dynamic superposition retrieval by taking a retrieval permission strategy as a constraint to obtain a plurality of refined result sets, carrying out descending order arrangement and integration on the refined result sets according to metadata timestamps, and outputting a communication data time series retrieval sequence. According to the method and the device, the technical problems of low retrieval efficiency and difficulty in accurately positioning the communication data caused by single semantic understanding in traditional communication data retrieval are solved, and the technical effects of accurate semantic understanding of the communication data, orderly and efficient retrieval and accurate positioning of the communication data meeting the intention of the user are achieved.

The invention relates to the technical field of computer testing, in particular to an incremental code coverage rate calculation method which comprises the following steps: executing a test case on a current version code, and collecting code test coverage information; comparing the baseline version code with the current version code to obtain a change code; based on a baseline version code, performing semantic equivalence analysis on the change code to obtain a logic change code, the semantic equivalence analysis being used for filtering change codes of non-logic factors; and calculating an incremental code coverage rate according to the logic change code and the code test coverage information, wherein the incremental code coverage rate comprises a change line coverage rate and a change link coverage rate. According to the multi-dimensional analysis method, coverage rate evaluation is more scientific and comprehensive, developers can better find test blind spots and potential defects, and therefore the software quality is improved.

The present application belongs to the technical field of software engineering, and specifically relates to a multi-function homologous vulnerability detection method based on function selection and semantic equivalent matching. The method comprises two stages of signature generation and vulnerability detection. The signature generation stage comprises: inputting a vulnerability repair patch, extracting a modification method and a key repair method therein; performing semantic equivalent replacement on the extracted key repair method, and generating signatures for the functions before and after the semantic equivalent replacement respectively; the vulnerability detection stage comprises: inputting a target detection software, pre-processing by using a clone detection software to reduce the search space; extracting functions in the target detection software and performing semantic equivalent replacement thereon, generating signatures for the functions before and after the semantic equivalent replacement respectively, and performing two-stage signature matching, and finally outputting whether there is a suspicious multi-method homologous vulnerability in the target project. The method can help identify multi-method homologous vulnerabilities in open source software, reduce the harm caused by homologous vulnerabilities, and improve the security of open source software.

The invention discloses a network traffic large model construction method based on heterogeneous federated learning, which comprises the following steps: analyzing, cleaning and formatting network traffic data, and constructing question and answer pairs; performing semantic equivalence reappearing on the public data set by using a large language model to generate an enhanced data set; performing architecture optimization on the generative basic model to replace an original output layer with a customizable network classification head; decoupling the teacher model into a simulator and an adapter; initializing a LoRA module for a simulator and an adapter of the student model; the client adaptively adjusts the LoRA rank according to local resources; the client only finely adjusts parameters of the LoRA module of the adapter and freezes other parameters; the server adopts a weighted stacking algorithm to aggregate the LoRA modules uploaded by the client; the method has good universality and expandability.

The invention provides a performance analysis method and device for a cross-architecture semantic equivalence instruction stream and a storage medium, and the method comprises the steps: taking a high-level language program as input, aiming at a plurality of target architectures, generating a multi-dimensional semantic equivalence instruction stream set from a high-level language form to an intermediate representation and then to an assembly code through a compiler; based on semantic equivalence instruction streams of code units with different granularities and compiler debugging information, establishing a mapping relationship between a target region in a source code and an intermediate representation and assembly codes, and forming a cross-hierarchy equivalence region fully-connected graph; and based on the semantic equivalence instruction stream set and the cross-level equivalence region fully-connected graph, automatically embedding a performance monitoring code at the boundary of the target region, and carrying out performance test on the elastic region. According to the method, the accurate performance test of the cross-architecture semantic equivalence instruction stream can be realized.

The embodiment of the invention discloses a data analysis method and device, electronic equipment and a storage medium, which are used for analyzing and aligning different field names in a plurality of historical data tables by using at least one of a large language model and a data mining algorithm, so that the data analysis efficiency is improved. The method comprises the steps that historical data is obtained, the historical data comprises a plurality of historical data tables, historical natural language query statements and historical structured query language SQL query statements corresponding to the historical natural language query statements, and each historical data table comprises at least one field name; the historical data are input into at least one of a large language model and a data mining algorithm, a target field alignment table is generated, and the target field alignment table comprises the mapping relation of different field names with semantic equivalence in multiple historical data tables; and according to the target field alignment table, updating the structured description information of the plurality of historical data tables, and generating updated structured description information of the plurality of historical data tables.

The invention discloses a big language model agent personalized response generation method and device, and relates to the technical field of personalized recommendation. Comprising the steps of obtaining a preference set from historical interaction behaviors, and constructing a group of preference chains for each preference based on a thinking chain; rewriting the interaction behavior at the current moment based on the generation function to obtain a plurality of semantic equivalence versions and an enhanced interaction set; determining the fine-grained alignment score of each group of preference chains and enhanced interaction sets and the relative strength of each preference based on a scoring function; taking the interaction behavior at the current moment and the response prefix with the initial state being null as basic input, and constructing a prompt template for each group of preference chains; and carrying out weighted summation on the probability distribution of the next token according to the relative preference intensity to obtain personalized target distribution, carrying out word-by-word sampling according to the personalized target distribution to generate the next token, splicing the tokens into a response prefix, updating the response prefix, and outputting a complete personalized response until the response is finished.