32 results about "Context sensitivity" patented technology

The invention discloses a mandatory access control method and device based on a process function context, and the method comprises the steps: collecting a security context associated with a system call initiated by a target process, so as to generate a standardized object description; mapping the object description into a target function classification identifier, so as to obtain a process function context view of the target process according to the target function classification identifier; constructing a target decision key for access decision based on the current policy era, the qualifier, the function classification identifier, the view identifier of the process function context view and the isolation domain abstract; and querying the multi-level cache according to the target decision key to determine a matched target access decision. Therefore, context-sensitive judgment and cross-component consistency taking the functional context as the center are realized.

The invention relates to the field of multi-modal neural network model compression, and particularly discloses a multi-modal model compression and distillation method and system based on causal reasoning, and the method comprises the steps: constructing a comprehensive causal discovery module, identifying a causal dependency relationship among the multi-modal features through information theory measurement, Granger causal analysis and intervention-based verification; executing an adaptive compression engine, and performing pruning, mixing precision quantification and low-rank decomposition based on a causal relationship; a cross-modal distiller is applied, and multiple loss function combinations are adopted to maintain the relationship between modals; and implementing a dynamic optimizer to carry out hardware perception and context-sensitive reasoning optimization. According to the method, the compression decision is guided through causal reasoning, the high compression rate is achieved while the key causal path is kept, and the deployment problem of the multi-modal model in the resource-constrained environment is effectively solved.

The application relates to the field of multi-modal neural network model compression, and specifically discloses a multi-modal model compression and distillation method and system based on causal reasoning, which comprises the following steps: constructing a comprehensive causal discovery module to identify the causal dependence relationship among multi-modal features through information theory measurement, Granger causality analysis and intervention-based verification; performing an adaptive compression engine to perform pruning, mixed precision quantization and low-rank decomposition based on the causal relationship; applying a cross-modal distiller to maintain the relationship among modes by using a variety of loss function combinations; and implementing a dynamic optimizer to perform hardware perception and context-sensitive reasoning optimization. The application guides the compression decision through causal reasoning, realizes high compression rate while maintaining key causal paths, and effectively solves the deployment problem of multi-modal models in a resource-limited environment.

The invention relates to the field of industrial data processing, and particularly discloses an industrial data anomaly detection method and system.The industrial data anomaly detection method comprises the steps that firstly, an alarm event is responded to capture a real-time context snapshot representing a physical world state, and after preliminary vector retrieval is completed, a set of multi-dimensional fidelity scoring mechanism based on risk self-adaptive gating is introduced; the mechanism firstly evaluates the potential influence of knowledge, dynamically generates a situation sensitivity index and a collection threshold value on the basis of the potential influence of the knowledge, and then achieves strong override on low-situation consistency knowledge, especially high-risk knowledge through an indexed gating factor. Finally, only the verified knowledge set passing the strict verification is constructed into a high-fidelity cue word and input into a large language model, it is ensured that a generated diagnosis report is established on the basis of high-quality knowledge which is matched with the current working condition, reliable in source and subjected to risk adaptation evaluation, and safe, credible and traceable diagnosis is achieved.

A machine learning-based method for accelerating a detection and disposition of content anomalies in a target digital artifact includes identifying, by one or more computers, a digital artifact underpinning a digital artifact assessment request; detecting, via the one or more computers, a plurality of content deviations in the target digital artifact based on a context-sensitive artifact assessment protocol obtained from the context-sensitive artifact assessment protocol repository; identifying, via a digital artifact assessment user interface, a sequence of one or more inputs corresponding to a rejection of a first subset of the plurality of content deviations; and based on identifying the rejection of the first subset of the plurality of content deviations: computing, via the one or more computers, a system-generated adaptation proposal for each content deviation underpinning the first subset of the plurality of content deviations based on machine learning-derived policies underpinning the context-sensitive artifact assessment protocol.

The invention provides a medical scene optimization-oriented OCR neural network based on a lightweight CRNN architecture, and aims to significantly improve the medical text recognition precision while reducing the calculation overhead. The core innovation of the method is that on one hand, a lightweight convolution feature extraction module is introduced, and the model scale and reasoning energy consumption are effectively compressed; on the other hand, a dictionary constraint mechanism in the medical field is fused, medical terms and semantic structure priori knowledge are utilized, and the recognition capacity for professional symbols, abbreviations and context sensitive content is enhanced. According to the architecture, high accuracy is kept, meanwhile, the requirement for computing power is remarkably reduced, efficient deployment can be achieved on medical edge devices (such as portable terminals and basic diagnosis and treatment devices) with limited resources, feasible technical support is provided for large-scale application of medical OCR in scenes such as mobile medical treatment, basic health services and intelligent diagnosis and treatment terminals, and the application prospect is wide. Important clinical and industrialization values are realized.

This invention discloses a method, device, and medium for predicting public transport carbon emission factors per passenger. The method includes: constructing bus stop-level, bus link-level, and bus route-level influencing factors; using these factors as input factors, processing them through a routing mechanism, calculating the routing score for each input factor, and then assigning each input factor to a corresponding expert sub-model to predict the corresponding public transport carbon emission per passenger. The method also involves identifying key factors for public transport carbon emissions and / or determining the coupling effect between pairs of input factors. Context-sensitive frequency and call balance are introduced to evaluate the local and global load of the expert sub-models, and the impact of expert switching on the prediction output is quantified through expert perturbation sensitivity.

Device for the secure hybrid execution of context-aware Large Language Model (LLM) deployments, comprising: an edge execution unit (101) configured to execute at least a first part of an LLM inference process on a local device; a cloud inference processing unit (102) configured to execute at least a second part of the LLM inference process on a remote computing infrastructure; a context sensitivity analyzer (103) configured to analyze an input query and associated context data and generate a sensitivity classification for said context data;a hybrid execution orchestration engine (104) that functions with the context sensitivity analyzer (103), the edge execution unit (101), and the cloud inference processing unit (102), wherein the hybrid execution orchestration engine (104) is configured to dynamically split the LLM inference process into an edge-executed part and a cloud-executed part, based on at least the sensitivity classification; a secure communication interface (105) configured to transmit intermediate inference outputs between the edge execution unit (101) and the cloud inference processing unit (102) over an encrypted communication channel; a TEE module (106) integrated into the edge execution unit (101) to perform at least one privacy-relevant inference operation in an isolated, hardware-protected environment;a key management and encryption controller (107) configured to generate, store, distribute, and rotate cryptographic keys and to encrypt and decrypt intermediate inference outputs transmitted between the edge execution unit (101) and the cloud inference processing unit (102); a policy enforcement and access control unit (108) configured to validate user authorization and enforce execution policies that define whether at least some of the context data is restricted to local execution;and an audit logging and threat monitoring module (109) configured to record inference execution events, security decisions, policy enforcement events and communication events and detect anomalous behavior related to the LLM inference process, wherein the device securely generates an output response to the input query by performing context-aware LLM inference while preventing sensitive context data from being disclosed to unauthorized external environments.

This invention relates to the field of enterprise office information security technology, and discloses a solution for office needs based on AI intelligent agents, including: context-sensitive semantic analysis, dynamic graph risk reasoning, multi-stage feature generation and dynamic learning, multi-level security policy generation, and feedback closed-loop optimization. Compared with existing technologies that rely solely on static keyword matching, lack dynamic risk reasoning and flexible permission adjustment, especially in handling complex conditions such as cross-network segment sharing, code names and coded messages, and abnormal operations, this invention addresses the technical problems of accurately identifying access to sensitive information and performing corresponding multi-level processing on sensitive access by using multi-stage feature modeling, path segmentation risk coding, context weighting, and improved XGBoost dynamic learning with AI intelligent technology. This achieves accurate labeling of sensitive information, forward prediction of leakage risks, and intelligent identification and control of permissions, thereby improving information security and the level of office intelligence.

A system and method for privacy-enhanced social networking may include dynamically managing user privacy and identity through multiple user profile aliases by defining each alias with a specific subset of user information and context parameters, such as location, proximity to other users, or participation in events. The system may dynamically select and display the appropriate alias based on the user's current context, thereby ensuring context-sensitive visibility of user information. Features include geofencing, proximity-based networking, and event participation tracking, supported by advanced system architectures such as graph databases and machine learning models. Additional functionalities include privacy-enhanced messaging with facial recognition for recipient verification, anonymous gifting, and automated will execution. A map interface and real-time adaptability further enhance user engagement and privacy management, fostering secure, meaningful interactions while maintaining robust privacy controls for applications in professional networking, community events, and confidential communications.

The invention discloses a demand-driven reusable binary pointer analysis method, which is used for determining a pointing set and an alias of a given target pointer variable, and comprises the following steps of: S1, recovering a pointer type in a process through a use mode of the variable, and extracting an instruction related to pointer calculation; s2, on the basis of the pointer representation and an execution result of the modeling instruction, abstract interpretation is carried out in the process, a pointer representation set of all pointer variables is obtained, and the result is stored in an intermediate result library; s3, on the basis of an analysis result in the process, determining a to-be-analyzed function as required, and obtaining different pointer representations of the target pointer variable in different functions; and if a certain function is not analyzed, the in-process analysis of the step S1 and the step S2 is triggered for the function. According to the method, a context sensitive strategy and a flow sensitive strategy are realized, the balance of analysis efficiency and precision is realized, and inter-process on-demand pointer analysis can be efficiently carried out.

The application discloses a method and device for enforcing access control based on process function context, wherein the method comprises: collecting a security context associated with a system call initiated by a target process to generate a standardized object description; mapping the object description to a target function classification identifier to obtain a process function context view of the target process according to the target function classification identifier; constructing a target decision key for an access decision based on a current policy epoch, a qualifier, a function classification identifier, a view identifier of the process function context view and an isolation domain digest; and querying a multi-level cache according to the target decision key to determine a matching target access decision. Thus, context-sensitive decision based on a function context is realized, and cross-component consistency is achieved.

A method of performing a data search in a data source by which an operator of a data search pipeline is just-in-time optimized and compiled, using an operator optimization module which optimizes and compiles an intermediate representation of the operator, considering runtime information, and optimization rules, to produce an operator that is optimized for the data search being performed. The method can be applied with one operator or with many operators applied in any sequence or tree structure according to a query plan, as determined by runtime information and optimization rules.

According to the context sensitive pointer analysis method provided by the invention, dependency and access relationships among modules are analyzed through keywords in the modules, modeling is carried out according to the dependency and access relationships, and a method calling statement for calling an interface is analyzed, so that the problem of completeness loss caused by the fact that interface access among the modules cannot be captured in a traditional pointer analysis method is solved. Secondly, an algorithm is designed, so that construction and updating of an object distribution graph and a module depth graph are completed in the statement analysis process, the pointer analysis efficiency is effectively improved, finally, key information is selected based on module depth information stored in the module depth graph to construct a context, the number of elements in the context is limited to be 1, and the algorithm is more efficient. Unnecessary analysis context elements are removed, and high precision of an analysis result is kept, so that the efficiency of an analysis process is remarkably improved.

The application relates to a data-to-text generation method and device based on dynamic division and electronic equipment. The method comprises the following steps: obtaining structured data, constructing a data-to-text generation network based on dynamic division, the network comprising an encoder and a decoder; the encoder is used for encoding the input to obtain a context-sensitive vector space; in the decoder, a planner and a generator are used alternately to perform data division and text generation, so that the planner can dynamically adjust the data division according to the generated content; an unsupervised training strategy is used to train the network to obtain a data-to-text generation model; the to-be-tested structured data is obtained and input into the model to obtain the text generated by the data. In the method, the planner can adjust the data division in a timely manner according to the generated text, and the unsupervised training network method can eliminate the dependence of the model on standard data division, thereby reducing the data labeling cost.

A system and method for privacy-enhanced social networking may include dynamically managing user privacy and identity through multiple user profile aliases by defining each alias with a specific subset of user information and context parameters, such as location, proximity to other users, or participation in events. The system may dynamically select and display the appropriate alias based on the user's current context, thereby ensuring context-sensitive visibility of user information. Features include geofencing, proximity-based networking, and event participation tracking, supported by advanced system architectures such as graph databases and machine learning models. Additional functionalities include privacy-enhanced messaging with facial recognition for recipient verification, anonymous gifting, and automated will execution. A map interface and real-time adaptability further enhance user engagement and privacy management, fostering secure, meaningful interactions while maintaining robust privacy controls for applications in professional networking, community events, and confidential communications.

This invention discloses a knowledge graph-based method and system for professional interaction within the hotel industry, relating to the fields of knowledge graphs and computing. The method includes: semantically parsing professional interaction data in the hotel industry to form industry semantic structure information; performing association matching in the hotel industry knowledge graph based on the industry semantic structure information to obtain a set of knowledge nodes corresponding to the industry semantic structure information; generating professional interaction context sensitivity parameters based on the knowledge node set and the industry semantic structure information; constructing knowledge association paths in the knowledge node set based on the professional interaction context sensitivity parameters to form a set of knowledge reasoning paths corresponding to the professional interaction; and integrating the professional interaction content based on the knowledge reasoning path set to generate structured professional interaction results. Through semantic parsing and graph consistency, the interaction content description and associated information are uniformly assembled into professional interaction results to support consistent collaboration in industry knowledge retrieval.

An augmented reality (AR) headset is disclosed that provides a wearer with spatial, system, and temporal contextual information of a surgical robotic system to guide the wearer in configuring, operating, or troubleshooting the surgical robotic system before, during, or after a surgical procedure. Spatial contextual information can be rendered in a coordinate system of the AR headset to display a spatially fixed 3D generated virtual model of the robotic arms, instruments, table, and other components of the surgical robotic system that matches the actual position or orientation of the surgical robotic system. The AR headset can communicate with the surgical robotic system to receive real-time status information of the components of the surgical robotic system. The AR headset can use the real-time status information to display context-sensitive user interface information, such as reminders, suggestions, visual or audio cues about maneuvering the robotic arms and table to their target positions and orientations or for troubleshooting purposes.

The application discloses a design parameter self-adaptive adjustment method based on intention perception and implicit interaction and related devices, wherein the method comprises the following steps: obtaining explicit operation data and implicit design behavior data of user operation in the process of user operation canvas; inputting the explicit operation data and the implicit design behavior data into a heterogeneous graph neural network model to output a context embedding vector; activating a context-sensitive rule subset in a parameter knowledge base according to the context embedding vector; constructing a dynamic parameter relationship hypergraph of the context-sensitive rule subset by using a hypergraph neural network model; when the user modifies any parameter, inputting the parameter relationship hypergraph and the user intention signal contained in the context embedding vector into a lightweight conditional diffusion network model to generate an optimal control combination scheme, and outputting a parameter completion suggestion according to the optimal control combination scheme. The application significantly improves the ability of the input component to process complex parameter relationships, and reduces the learning cost of the user through intelligent input auxiliary functions.

The invention provides an event site detection method and system based on retrieval enhancement generation and multi-view verification, and relates to the technical field of event site detection methods.The event site detection method comprises the steps that to-be-detected texts are obtained and numbered, each text is input into a Qwen3 model, a position vector of each word is extracted to serve as a semantic representation vector of each text, and the semantic representation vector of each text is obtained; obtaining a graph semantic vector at the same time, and calculating a final matching score of the to-be-queried statement in each to-be-detected text; and extracting the matching scores of the to-be-queried statements in each text, sorting the matching scores, and selecting the first three texts with the highest matching scores as evidence texts. According to the method, context sensitive semantics of a text are extracted through a position vector of Qwen3, and meanwhile, graph structure representation of AMR analysis is introduced, so that the capture degree of a logic structure is improved.

This invention discloses a context-sensitive memory vulnerability detection method and system for Rust programs. By implementing context-sensitive alias analysis and taint analysis, it detects dangling pointers, use-after-free, and double-free memory vulnerabilities in the unsafe region of Rust compilation units (library crates, binary crates). This invention innovatively employs context-sensitive alias analysis to increase analysis accuracy and reduce the false negative rate.

The present application relates to the technical field of data mining, and more particularly to a kind of enterprise financial data analysis management cloud platform system and its programming method.The method comprises the following steps: collecting multi-source enterprise financial data, and carrying out analysis pre-classification and compression coding, to obtain time data acquisition event stream;Based on time data acquisition event stream, context-sensitive semantic alignment is carried out, to obtain high-dimensional semantic tensor;According to high-dimensional semantic tensor, external economic entity semantic constraint is carried out, to obtain financial entity relationship graph;The logical relationship between the nodes of financial entity relationship graph is structured chain encapsulation and consensus verification, to obtain encrypted verification financial chain package;The encrypted verification financial chain package is programmed logic constraint, to obtain pre-constructed logic execution unit set, and the pre-constructed logic execution unit set is uploaded to enterprise private cloud platform, to execute platform building task.The present application enhances the intelligent level and management efficiency of enterprise financial analysis management.

This invention discloses a traffic classification method, electronic device, and medium based on session context-sensitive BERT, comprising: acquiring network traffic data; performing session segmentation on the network traffic data to obtain several traffic sessions; tokenizing each traffic session to obtain a session byte sequence; extracting the packet length sequence of all data packets between multiple traffic sessions within a time window and performing multi-granularity temporal aggregation to obtain an inter-session packet length sequence; constructing a session context-sensitive BERT model; the processing includes: encoding the session byte sequence and the inter-session packet length sequence to obtain a first embedding vector and a second embedding vector; using the concatenated first embedding vector and second embedding vector as training samples to construct a joint loss function, thereby pre-training the model; fine-tuning the pre-trained model for downstream traffic classification tasks; and using the fine-tuned model to perform downstream traffic classification tasks.

The invention discloses an intelligent code vulnerability automatic detection system for basic software development, and belongs to the technical field of software security and static analysis. The system comprises a source code preprocessing and standardization module, a multi-level composite abstract syntax tree construction module, a domain knowledge enhanced vulnerability rule knowledge base, a graph neural network-based context sensitivity analysis module and a symbolic execution and constraint solution guide module. A vulnerability association and root cause positioning module; and a feedback-driven adaptive optimization engine. According to the method, enhanced code representation fusing grammar, semantics and specific field features is constructed, deep semantics, complex data streams and control streams of codes are modeled in combination with a deep learning model, and high-precision and context-sensitive automatic detection of various vulnerabilities such as memory security, concurrent security and logic defects is achieved.

The invention provides a knowledge base intelligent retrieval method based on a large language model, which comprises the following steps: integrating user query and at least three rounds of time sequence contexts, and adopting standardized text preprocessing to generate structured input; using a pre-training language model and a position awareness attention mechanism to jointly encode dialogue content to form a context sensitive semantic vector; a two-channel vector updating structure is constructed, and a knowledge vector state is jointly controlled through semantic consistency comparative learning and attention distribution evolution detection; a differentiable memory gating unit is adopted to update and retain dynamic vectors; a vector backtracking mechanism is designed to improve multi-round semantic coherence, multi-modal user feedback signals are collected, and continuous online optimization of gating parameters is supported.

Systems and methods address the inefficiencies in document creation and management within large, globally dispersed organizations. The system leverages multiple independent cloud-based Generative AI (GenAI) models, each specialized and context-sensitive, to dynamically generate and contextualize documents tailored to specific regional and contextual requirements. A global repository meticulously catalogs and annotates documents with detailed metadata, enabling precise retrieval based on user-specific search criteria. The system employs a denormalization slant to generalize search queries, facilitating global context searches and affinity mapping to ensure relevance. A Primary Context Controller Hyper Model orchestrates the confederation of independent GenAI models, refining and enhancing documents to meet precise user needs while adhering to local nuances. Automated, rule-based approvers ensure instant validation and quality control, while a feedback loop mechanism continuously improves the models' accuracy and relevance. This decentralized, scalable approach reduces computational burden and costs, providing a robust, efficient solution for high-quality, context-aware document generation.

The embodiment of the invention discloses a semantic recognition method and device, equipment, a medium and a product. The semantic recognition method comprises the steps of obtaining multi-modal input data corresponding to data to be recognized; the multi-modal input data comprises text data and image data; performing alignment processing on the context sensing vector, the text data and the image data through a dynamic encoder and a cross-modal alignment mechanism to determine text dynamic features, image dynamic features and context features; the context feature is determined based on the to-be-identified data; performing weighted fusion on the text dynamic features, the image dynamic features and the context features through a cross attention mechanism and a gating weighted fusion mechanism to determine a cross-modal joint feature vector; and determining a semantic recognition result of the to-be-recognized data based on the cross-modal joint feature vector. According to the technical scheme, the problem that a traditional alignment method cannot adapt to data distribution changes and semantic hierarchy differences is solved, and cross-modal semantic matching with fine granularity, context sensitivity and dynamic evolution is achieved.

The invention discloses a Chinese resume unstructured privacy entity recognition method, which comprises the following steps that: a word segmentation layer solves the problem of text coding and label alignment, and adopts a subword-based segmentation strategy and combines a filling and attention mask mechanism; the feature coding layer uses a RoBERTa pre-training model to extract context-sensitive feature representation, and the training efficiency is improved through the LoRA technology; the label classification layer realizes mapping from features to label space; the sequence optimization layer introduces a conditional random field (MCRF) with a mask constraint matrix to model a global dependency relationship of a label, so that the recognition capability of a model on an entity boundary is improved; and the decoding layer adopts a Viterbi algorithm to generate a final entity recognition result from a prediction result. According to the method, the representation capability of the deep learning model and the structured prediction capability of the traditional sequence model are combined, the accuracy and stability of privacy entity recognition of the unstructured text in the Chinese resume are improved, and the method is suitable for data privacy protection and key information extraction scenes of the Chinese resume.