106 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 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 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 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.

The invention provides a source database storage process migration method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the grammar analysis of an original storage process in a source database, and generating a corresponding abstract syntax tree; based on a preset rewriting rule set, performing semantic equivalent transformation on the flow control syntax structure in the abstract syntax tree to generate a transformed abstract syntax tree; and generating a target script in a target execution environment from the converted abstract syntax tree, executing the target script in a script running engine, dynamically generating a standard SQL statement corresponding to the business logic defined by the original storage process, and submitting the standard SQL statement to one or more target data platforms for execution. And the business logic of the original storage process is reproduced. By introducing the target script as the intermediate execution layer, the flexibility of the migration scheme is greatly improved, so that the migration scheme is suitable for various big data platforms without the native storage process capability.

Examples detect equivalent subexpressions within a computational workload. Examples include converting a query plan tree associated with a first subexpression into a matrix. The first subexpression is a portion of a database query from the computational workload. Each node in the query plan tree is represented as a row of the matrix. The matrix is converted into a first vector. The first subexpression is determined to be equivalent to a second subexpression by comparing the first vector to a second vector associated with the second subexpression. The comparison includes computing a distance between the first and second vectors that is lower than a distance threshold. The computational workload is modified, based on the determining, to perform the first subexpression and exclude performance of the second subexpression as duplicative.

The invention relates to the technical field of data management, and particularly discloses an efficient management method and system for a data directory in a data space, and the method comprises the steps: carrying out the incremental updating of metadata information, data product registration information and cross-node data directory information related to the data directory to a distributed data storage system; constructing a full-text index and a vector index of metadata in the distributed data storage system, and matching a vector generated based on the request content with the full-text index or the vector index to obtain a retrieval response result; marking a semantic equivalence data set based on the semantic vectors of all the data sets, constructing a data set node graph based on all the semantic entity lists, and outputting a semantic equivalence field list based on the semantic equivalence data set and the data set node graph; determining an associated recommendation result of the retrieval response result based on the semantic equivalence field list, and synchronously returning the retrieval response result and the associated recommendation result; data directory management and efficient utilization in a distributed heterogeneous cross-domain data space are supported.

The disclosed embodiments describe a method, system, and computer-readable medium for generating a training dataset for training a model in the field of natural language processing involving receiving a set of input samples and performing a rephrasing operation to produce new versions of the set of input samples, where the new versions preserve semantic equivalence as the set of input samples but have different phrasing. A dataset of generated versions of the input samples is generated using a generative Language Learning Model (LLM), all entity references present in the generated versions of the input samples are labeled, and the generated versions of the input samples and their corresponding labeled versions to form an expanded labeled dataset are aggregated.

The invention relates to a short message copywriting sensitive word recognition and analysis method based on artificial intelligence, and provides a deep semantic modeling and cross-modal semantic alignment mechanism based on multi-modal data aiming at the technical problem of realizing rich media display and text content synthesis of marketing short messages under compliance requirements. And combining the pre-trained language model and the convolutional neural network to respectively extract high-dimensional semantic features of the text and the image, and optimizing semantic consistency through an attention mechanism and a contrast loss function. Furthermore, a dynamic sensitive word recognition and avoidance strategy is introduced into the system, sensitive expressions are effectively filtered, and compliance semantic equivalence content is generated through a text reconstruction model when necessary. And finally carrying out format fusion on the content and the image to generate a rich media short message which can be sent in compliance. According to the scheme, semantic accuracy and content compliance of multi-modal information fusion are improved, it is ensured that marketing short messages meet legal compliance requirements, and the marketing short messages have good user attraction and a good transmission effect.

The invention discloses a WebShell malicious behavior fine grit method based on an abstract syntax tree and de-obfuscation, which comprises the following steps: inputting a WebShell suspicious script file into a computer system, and quickly identifying and normalizing the suspicious script through file format screening, feature matching and file standardization to obtain a normalized script; executing a de-obfuscation operation on the standardized script based on a large language model to obtain a simplified semantic equivalence code; analyzing the code to generate an abstract syntax tree, extracting key features through a multi-granularity feature extraction mechanism, and capturing multi-dimensional context dependency; and extracting multi-dimensional feature vectorization, performing feature fusion, inputting a deep neural network combined with ResNet, BiLSTM and Transform, and outputting a WebShell malicious behavior detection result and a malicious behavior fine-grained classification result. The WebShell attack detection method can solve the problems of low detection accuracy and high false alarm and missing report rate of an existing detection method when facing highly hidden, deformed and confused WebShell attacks.

The invention provides an SMT solver rewriting system defect detection method based on rewriting space exploration, and aims to solve the problems that in the prior art, an SMT solver rewriting system is insufficient in verification, test input is difficult to trigger a rewriting rule, errors are difficult to verify and the like. The method comprises the following steps: 1) collecting and automatically verifying rewriting rules in an SMT solver, and constructing a high-quality rule base; 2) performing structure transformation on the existing test formula by adopting mimicry variation to enable the existing test formula to be matched with a rewriting rule mode, thereby activating more rewriting opportunities; 3) constructing an equivalence graph based on a generative equivalence saturation technology, and systematically generating semantic equivalence but structural diversity formula variants; and 4) inputting the original formula and the variant formula into an SMT solver, comparing the solving result with the performance behavior, and detecting potential defects. The invention also provides corresponding system implementation, and supports an automatic test process, rule management, formula generation and defect recording.

The invention relates to the technical field of body intelligence, in particular to a body intelligence attack detection method based on a constraint function, mainly solves the technical problems of insufficient granularity, poor real-time performance and insufficient robustness of the existing attack detection method, and comprises the following steps: S1, selecting key points; s2, generating a constraint function; s3, analyzing the images where the key points are located by adopting a credible multi-modal model, and generating task fingerprints; s4, comparing the constraint function with the task fingerprints, and calculating an output difference; semantic equivalence measurement is introduced; a tolerance value is introduced, if the output difference exceeds a threshold value, it is judged that the constraint function is attacked, and the system stops task execution. According to the method, the detection accuracy exceeding 92% can be realized in various attack scenes; the problems of false report and missing report caused by insufficient rule coverage or large language model output diversity can be avoided, and the robustness is high; and the execution safety and the real-time performance of the body-equipped intelligent agent in a complex and open environment can be obviously improved.

This application discloses a method, system, and device for detecting malicious code variants using control flow graph semantic matching, belonging to the field of computer network security technology. It includes: disassembling the target executable file, constructing a control flow graph, and extracting basic block semantic features to generate semantic vectors; obtaining semantic equivalence classes based on semantic vector clustering, identifying main anchor points, and constructing a main anchor point skeleton graph; performing semantic compression and noise processing with the anchor point skeleton as constraints, and obtaining a semantic core graph through topological verification; hierarchically normalizing the semantic core graph and the main anchor point skeleton graph to generate a composite hash signature; and matching the composite hash signature with a malicious code family signature library to complete variant determination. The above scheme, through the main anchor point constraint compression and normalization process, resists structural obfuscation disturbances, improves the consistency of homologous variant identification, and has advantages such as strong robustness, high efficiency, and good interpretability. It is suitable for detecting malicious code variants using control flow graph semantic matching in complex obfuscated environments.

This invention relates to the field of data management technology, specifically to a method, apparatus, computer equipment, and storage medium for multi-source data consistency management. The method includes: acquiring data records from at least two source systems to generate change events; converting the change events into evidence records containing original values ​​and contextual information based on a semantic rule registration library, and generating a standardized value view; performing entity parsing based on the standardized value view to obtain a set of candidate entity matching relationships, and outputting a credible candidate set through semantic equivalence transformation stability testing and evidence conflict measurement; performing constraint consistency inference based on the credible candidate set and the semantic rule registration library to generate a consistency state snapshot and a correction action plan; generating a synchronous execution rule set based on the correction action plan and performing a pre-rehearsal acceptance test; executing correction and re-collecting verification when successful, and rolling back when unsuccessful. This invention improves the traceability of cross-source consistency determination and the verifiability of correction execution, and reduces the risk of erroneous corrections.

The application discloses a kind of semantic equivalent text generation methods based on familiar word constraint. Among them, the method includes: obtaining user familiar word set, and based on the familiar word set, construct structured limited expression space, for limiting the range of vocabulary and expression structure range that can be used in text generation process;Target text is carried out semantic structure analysis, and the target semantic structure representation is generated;In the structured limited expression space, semantic matching solution is executed, and text generation problem is converted into limited semantic solution problem, at least one expression path that satisfies the target semantic structure representation is constructed under the condition of the limited expression space;Expression candidate set is generated based on the expression path;The semantic consistency determination is carried out to the expression candidate set, and the semantic equivalent expression variant set that satisfies consistency condition is output.The application solves the technical problem that existing text generation technology is difficult to structure solution for target semantics within limited vocabulary range.

The invention discloses a large model reasoning stability enhancement method and system in a text scene, and relates to the technical field of knowledge enhancement deep learning, and the method comprises the steps: building a structured text knowledge graph based on a text knowledge base, and generating a dynamic text knowledge embedding matrix through a graph attention network; inserting a text knowledge gating cross attention module into a decoder selection layer of the pre-trained large model, taking the text knowledge gating cross attention module as an external knowledge source, obtaining a knowledge enhanced hidden state after gating fusion, and constructing a transformation model; semantic equivalent perturbation is carried out on an input text to obtain a perturbation sample, the perturbation sample is input into the transformation model in parallel to obtain extraction probability distribution, and divergence and gradient direction consistency loss between two distributions are calculated; and combining cross entropy loss and gradient direction consistency loss to train and transform the model, and updating parameters to convergence to obtain a final large model. According to the method, the fact consistency of output can be improved, common optimization of knowledge guidance and stability constraint is realized, and the result is accurate and reliable.

This invention discloses a detection method based on a large language model, belonging to the field of natural language processing technology. The method includes: collecting and preprocessing the original detection text; generating detection prompts by constructing dynamic contextual anchors based on the preprocessed data through semantic scoring and redundancy suppression; obtaining the main original detection response through the large language model based on the detection prompts, constructing semantically equivalent perturbation text, and evaluating the stability of the large model's detection response; parsing the response conclusions and calculating the perturbation consistency confidence score; making a detection judgment based on the perturbation consistency confidence score and outputting structured detection results; and performing self-calibration of the confidence score threshold using the structured detection results. This method not only significantly improves the accuracy, interpretability, and robustness of detection but also effectively enhances resistance to user rewriting avoidance behavior and reduces the risk of detection errors caused by missing context or superficial wording changes.

This invention relates to the field of natural language processing technology, and more particularly to a method and system for language transfer recognition based on cultural semantic embedding. The method includes the following steps: acquiring source language data and target language data; performing semantic isomorphic centering on the source and target language data to obtain semantically equivalent data; decoupling pragmatic features based on the semantically equivalent data to obtain pragmatic feature data; performing contrastive cultural latent space embedding based on the pragmatic feature data to obtain bilateral cultural representation data; performing cultural modality affine transformation based on the bilateral cultural representation data to obtain cultural difference data; performing implicit distortion pattern discrimination based on the cultural difference data to obtain cultural semantic distortion data; and generating transfer results based on the cultural semantic distortion data to obtain language transfer result data. This invention decouples pragmatic features under semantic equivalence constraints and achieves language transfer recognition through cultural semantic embedding and modality affine modeling.

The application discloses a kind of based on semantic context's maintenance information aggregation search method and system, it is related to natural language processing technical field;After text analysis to search request, based on the maintenance semantic context extraction of preset static maintenance semantic label library obtains dynamic context;After the structured organization construction of dynamic context and search request, multiple semantic equivalent rewriting is carried out to obtain multiple rewritten search requests;Based on the scheme maintenance result search of preset multi-source maintenance knowledge base to all rewritten search requests obtains search result set;Based on dynamic context, search result set is sorted, and the first preset search result is obtained as final output.Analysis user natural language request, then rely on static maintenance semantic label library to complete dynamic context, then by structured fusion and multi semantic rewriting search, aggregate multi-source knowledge without omission, to quickly obtain the accurate scheme of adaptation, improve search accuracy while also greatly improve search efficiency.

The application provides a large model NL2SQL evaluation method and device based on database exploration, and relates to the technical field of data processing. The method constructs the semantic correctness evaluation process of NL2SQL as a multi-step decision process driven by a large model agent. The large model evaluation agent module generates a probing query by calling a read-only database execution tool module, iteratively updates the context according to the dynamically returned execution results, and outputs a semantic equivalence judgment. The present scheme changes the traditional "static text matching" evaluation paradigm to "dynamic execution verification". Through the closed-loop feedback mechanism of "proposing a hypothesis - executing verification - updating evidence", the evaluation bias caused by a single reference SQL is effectively solved, complex cases such as "accidental consistency of execution results but semantic error" or "inconsistency of execution results but semantic correctness" can be identified, and the execution accuracy of the evaluation and the robustness of the semantic discrimination are significantly improved.

The application discloses a large language model intelligent agent personalized response generation method and device, and relates to the technical field of personalized recommendation. The method comprises the following steps: 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 current moment interaction behavior based on a generation function to obtain a plurality of semantic equivalent versions and an enhanced interaction set; determining the fine-grained alignment score of each group of preference chains and the enhanced interaction set and the relative strength of each preference based on a scoring function; taking the current moment interaction behavior and the response prefix with an initial state of null as a basic input, and constructing a prompt template for each group of preference chains; performing weighted summation on the next token probability distribution according to the relative preference strength to obtain a personalized target distribution, word by word sampling to generate the next token according to the personalized target distribution, splicing the token into the response prefix and updating the response prefix until the complete personalized response is output.

The invention belongs to the technical field of network security, and particularly relates to a Web application firewall vulnerability detection method based on load variation injection. Comprising the following steps: constructing a unified attack grammar rule base covering multiple attack types; analyzing the original attack load into an abstract syntax tree; carrying out semantic reservation load variation based on the syntax tree to generate a semantic equivalent variation load; performing protection rule detection on the Web application firewall, analyzing a detection result and performing pruning operation on the syntax tree; a Monte Carlo tree search algorithm is utilized, a rule detection mechanism is combined, and a variation path with a high success rate is dynamically evaluated and preferentially tested so as to efficiently generate an antagonistic load capable of bypassing WAF detection; according to the method, the problems of attack failure, low black box test efficiency and insufficient coverage of a single attack type caused by semantic destruction in the prior art are solved, and the WAF rule base vulnerability can be systematically evaluated.

The invention belongs to the technical field of software engineering, and particularly relates to a homologous vulnerability detection method based on graph semantic enhancement matching. The method comprises the following steps: (1) a signature generation stage: inputting a vulnerability repair patch, constructing a standardized call graph through semantic equivalent transformation, extracting a key vulnerability path and a patch repair path in a patch modification function and generating a type code attribute graph as a semantic signature by combining cross-process stain tracking and an in-process dependency slicing technology; and (2) a vulnerability detection stage: inputting a target project, constructing a standardized target project call graph, obtaining candidate suspected vulnerability methods and sensitive statements and variables thereof, obtaining a code attribute graph of the target project by adopting a slicing method, and calculating similarity by adopting a graph semantic matching mode so as to judge whether the target project has homologous vulnerabilities or not. The method can help to identify the homologous vulnerability in the open source software, and improve the prevention and control capability of the open source software on the homologous vulnerability risk in the code reuse scene.

The invention discloses a semantic entropy-based big language model illusion detection method and device and a medium, and relates to the technical field of natural language processing and artificial intelligence. The method comprises the following steps: acquiring a plurality of answers generated by a large language model for the same target question under given sampling parameters; performing semantic equivalent clustering on the plurality of answers through a bidirectional implication relationship to obtain a semantic cluster; calculating semantic entropy according to the semantic cluster, and judging whether the answer is an imaginary content or not; and when the semantic entropy is higher than a preset threshold value, executing problem reconstruction and confidence coefficient consistency calibration, and comprehensively judging whether illusion appears or not. According to the method, semantic entropy calculation and a consistency calibration mechanism are fused, and the limitation of a single detection method is effectively overcome. Firstly, semantic clustering is carried out through a bidirectional implication relationship, the accuracy of semantic uncertainty measurement is improved, and the risk of misjudgment caused by expression difference is reduced.

The invention relates to the technical field of computers, in particular to a visual programming method and device and a readable storage medium. The method comprises the following steps: constructing a node graph of a double-layer connection system adopting separation of a data flow and a control flow to represent an algorithm logic; generating an abstract syntax tree which is semantically equivalent to the node graph, and establishing a bidirectional mapping relationship between the abstract syntax tree and the node graph; when the algorithm logic is executed, real-time memory state data including a call stack and a heap memory are captured through a tracking probe; and according to the data, dynamically generating a memory model visual view containing call stack and heap memory tracking. According to the method, the problems of unclear logic expression and opaque execution process in the prior art are solved, and the development and debugging efficiency is remarkably improved by intuitively presenting the dynamic execution process of the algorithm.

The invention discloses an intelligent question and answer method and system based on double-stage retrieval and generation. The method comprises: receiving a user query; vectorizing the questions through a mixed retrieval model, firstly retrieving in a standard knowledge base, and directly returning corresponding standard answers if semantic equivalence standard questions are found; otherwise, retrieving again in the policy original text knowledge base, and generating a final answer through a retrieval enhancement generation technology based on a retrieval result. Wherein the hybrid retrieval model is formed by carrying out weight fusion on the results of the original Embedding and the fine adjustment Embedding. According to the system, high-frequency standard questions and low-frequency complex questions are distinguished through a double-stage mechanism, the response efficiency and answer consistency are guaranteed, and meanwhile complex consultation needing policy original text reference can be processed. The hybrid retrieval model uniformly supports two-stage retrieval, takes into account general semantic understanding and domain optimization, and improves the retrieval quality and the system stability.