52 results about "Semantic memory" patented technology

The invention belongs to the field of artificial intelligence and robot navigation, and discloses a visual language navigation method and system based on a cross-task incremental semantic memory graph, and the method comprises the steps that an intelligent agent executes a zero-sample visual language navigation task in a continuous environment; performing cross-modal alignment on the natural language instruction and environment observation based on a multi-modal large language model, selecting candidate waypoints and updating task progress; a semantic memory graph is constructed and dynamically updated, wherein the semantic memory graph is used for structured storage and cross-task multiplexing of scene semantic information and a spatial topological relation sensed by an intelligent agent in historical tasks; and performing global path planning and local dynamic fine tuning based on the semantic memory graph. According to the method, the problem of task-by-task forgetting in a traditional method is solved, environment understanding and task reasoning capabilities are improved by constructing structured long-term memory, and navigation precision and robustness are optimized through a global-local collaborative strategy.

The invention relates to a multi-target tracking method based on dynamic semantic focus migration, and belongs to the field of intelligent vehicle target tracking. The method fuses image and language information, and comprises the following steps: firstly, respectively extracting visual features of an image and semantic features of a language instruction, obtaining a candidate target frame, constructing a semantic memory pool, and generating a semantic focus migration rule matrix describing a user attention evolution trend; secondly, guiding visual features and semantic features to perform spatial alignment based on a semantic focus migration rule matrix, and fusing visual and language information through a multi-head cross attention mechanism to obtain multi-modal fusion features; and finally, combining the fusion features and the semantic focus migration rule matrix with visual similarity to construct a multi-target association graph, optimizing a matching relationship by using a graph neural network, improving the tracking stability in complex scenes such as shielding and appearance mutation through robustness verification and strategy scheduling, and outputting a final multi-target tracking result.

The invention relates to the field of man-machine interaction, and discloses a man-machine interaction method and system based on a large language model, and the method comprises the steps: extracting original semantic features from the current input of a user; obtaining a context representation of the current dialogue; extracting a historical emotional state sequence from the multi-round dialogue historical record; generating a current user emotion representation vector based on the context representation of the current dialogue and the historical emotion state sequence; constructing a hierarchical memory structure based on the current user emotion representation vector and the context representation of the current dialogue; adjusting the attention degree of the historical dialogue content stored in the semantic memory component according to the hierarchical memory structure; and generating a response sequence through a large language model based on the adjusted attention degree, the hierarchical memory structure and the context representation of the current conversation, and outputting the response sequence to the user. According to the technical scheme, the problems of stiff switching and dialogue breakage during emotion turning of a traditional dialogue system are solved.

The invention provides a Suzhou dialect speech recognition system and method based on a tone track neural field, and the system comprises a tone track neural field module which is used for modeling the tone change of Suzhou dialects into a continuous space-time neural field; the bidirectional semantic memory network module comprises a forward prediction memory bank and a backward correction memory bank; the phoneme-font coupling error corrector is used for realizing polyphone disambiguation and homonym error correction by establishing association mapping between a phoneme sequence and a font sequence; the semantic entropy calculation module is used for evaluating the uncertainty of the recognition result; and the self-adaptive fusion decision module is used for generating a final recognition text. According to the invention, through an online learning mechanism, the system can continuously accumulate experience from actual use, automatically discover a new language mode and update an identification strategy. The self-improvement capability enables the system to adapt to the dynamic change of languages, and the performance is continuously improved along with the increase of the use time. Each use of the user helps the system to become more intelligent and accurate.

The invention discloses a memory optimization system for a large-scale language model, and the system comprises a dynamic semantic memory modeling module which is used for generating a context-aware memory vector through a bidirectional Transform, and supporting the increment to be updated to one hundred million entity records; the self-adaptive memory resource scheduling module dynamically allocates memory pool partitions based on inference load, high-frequency memory is stored in a working area, and quantitative compression is started in an archiving area; the memory attenuation and enhancement module adopts a two-factor attenuation mechanism to execute enhanced storage and regular pruning on the high-value memory; the cross-scene memory migration module is used for realizing cross-domain adaptation through a domain adapter and comparative learning; and the memory conflict resolution module identifies conflicts based on the rule base and adopts a multi-factor voting mechanism to resolve the conflicts. According to the method, the memory retention integrity and the retrieval efficiency are improved, the resource utilization is optimized, the cross-scene adaptation capability is enhanced, the memory conflict is eliminated, and the model reliability is improved.

The invention relates to an intention recognition system based on a multi-mode agent and data annotation, in particular to the field of artificial intelligence, according to the scheme, a semantic gap between different modes is effectively bridged by constructing a unified semantic projection space, the reliability of information of each mode is quantitatively and dynamically evaluated by using uncertainty, weighted fusion is performed on the basis, and the intention recognition accuracy is improved. A cross-modal semantic memory network and an online learning mechanism are innovatively introduced, so that the scheme not only can make more accurate intention inference by integrating historical experience and real-time perception, but also can continuously optimize own confidence assessment ability and a memory bank in an interaction process; therefore, the method has higher adaptability and robustness in the environment with noise, missing information and dynamic change.

The invention discloses a browser sidebar intelligent assistant interaction method and system, and relates to the technical field of man-machine interaction, and the method comprises the steps: obtaining webpage data according to webpage context information, obtaining interaction data from a user interaction record, and obtaining user emotion data through the interaction data; generating a task recommendation queue based on the webpage data, the interaction data and the emotion data by using a multi-dimensional reasoning method, and generating candidate interaction items according to the task recommendation queue and the user emotion data; receiving an instruction input by a user through the candidate interaction item, generating an analysis result in combination with the webpage data, executing a preset task corresponding to the analysis result, and obtaining task result data; and selecting a visualization template according to the task result data and the user emotion data to output a task result, and updating the user personalized configuration and the semantic memory library. The method is used for solving the problems that traditional browser assistant task recommendation lacks pertinence, and the matching degree of interactive experience and user states and preferences is low.

A system capable of reasoning, learning, and imagination. The system includes an input / output module, a reasoning and decision agent, and a knowledge management module including a deliberation agent and a semantic knowledge space. This system uses prior knowledge stored in memory for reasoning and decision-making. It learns new domain knowledge and user behavior throughout operation, making it an evolving system that adapts to the user's needs. The system reinforces knowledge stored in its semantic memory without user intervention and imagines new relationships between existing concepts to search for novel ideas until it reaches an epiphany. Several embodiments of the disclosed system can interact in a collaborative environment for cross-domain reasoning and brainstorming new ideas.

The computing system (10) monitors inference conditions of the generative model (74); detects a predetermined trigger condition (48) in the monitored inference conditions; and in response to detecting the predetermined trigger condition (48), merges a memory vector database (76) of the generative model (74) to extract semantic memories (66) from the vector database (76); updates the generative model (74) using the extracted semantic memories (66), and deploys the merged generative model (74). The predetermined trigger condition (48) can be at least one of a database size condition, an available memory size condition, a processor load condition, or a schedule time condition.

A system and method for implementing persistent cognitive computation through geometric representation augmented with holonomy-based experiential memory. The system encodes inputs into a curved latent manifold and maintains bounded sets of holonomy descriptors at each location, enabling two-component cognitive states comprising position and experiential context. Cognition occurs through holonomy-sensitive traversal where paths depend jointly on geometric structure and accumulated path-dependent constraints. Holonomy generators are created during traversal from prediction errors and constraint encounters, composed into consolidated descriptors, and undergo lifecycle management including reinforcement, decay, and irreversible export to residual constraint regions. This architecture escapes location-only representations by distinguishing cognitive states that occupy identical semantic positions but arise through different experiential histories. The system supports counterfactual reasoning through holonomy switching at fixed locations and preserves semantic memory as compressed transport deformation rather than stored trajectories, enabling scalable experiential learning where repeated patterns strengthen constraints while capacity remains bounded.

The invention relates to a construction scheme understanding degree visual guiding method based on structured semantic memory, which comprises the following steps of: converting a construction scheme editing and auditing process into an interactive, guidable and measurable structured cognitive cycle by constructing a dynamic task planning and project semantic map dual-core driving mechanism. The method has the beneficial effects that the problems of semantic understanding fragmentation, low man-machine interaction efficiency, task flow stiffness and the like in a traditional mode are effectively solved, deep fusion of transparent presentation of AI understanding degree and user accurate guidance is realized, and the intelligent level and cooperation efficiency of building scheme editing and examination are remarkably improved.

Mechanisms are disclosed for detecting one or more defects in a system In-system and externally observable events are received. Active state information is derived from the in- system and externally observable events. If a memory update is required based on facts from the semantic memory, an episodic memory and a working memory are updated. Actions required from a procedural memory are extracted based on the updated episodic memory and the updated working memory. A prompt is executed on a generative language model to produce a model reasoning output. Reproduction steps are derived from the episodic memory, based on a determination that a defect is present from the model reasoning output. A defect signal is generated and provided to a testing user.

The application discloses an intelligent terminal semantic understanding optimization method based on deep learning, and relates to the technical field of intelligent terminal semantic understanding. When a current short instruction is received, the method reads multi-source interface events in a preset time window and constructs a short-time interface state transition graph. Based on the short-time interface state transition graph, a local causal chain is extracted and an interface attribution clue sequence is formed. Double-granularity semantic coding and slot constraint decoding are performed on the current short instruction text, and cross alignment is performed with the interface attribution clue sequence to generate a task attribution candidate result. Path backtracking and consistency checking are then performed to determine a target attribution result. Finally, a call entry is determined according to the target attribution result and written into a short-range semantic memory cache. The application is suitable for multi-application switching, page jumping and short instruction continuous input scenes.

The application belongs to the technical field of time series data analysis, and discloses an online time series event analysis method and system based on heterogeneous memory fusion, which comprises the following steps: acquiring streaming time series data and preprocessing, extracting features from the preprocessed data to obtain time series feature sequences; long-term feature memory, event state probability memory and event-level semantic memory are constructed and fused to form a fusion memory; an event query representation is initialized, and the event query representation is enhanced through a short-term memory and a fusion memory in turn to generate a final event query representation; based on the final event query representation, an instantaneous event state and a candidate event instance at the current time are generated, the candidate event instance is de-redundant and cross-time deduplication processed, and a final event set is output. By constructing and fusing heterogeneous historical memories, the application realizes effective utilization of multi-granularity time series knowledge, thereby significantly improving the accuracy, stability and real-time performance of online event analysis under strict causal constraints.

The invention discloses a multi-round intelligent customer service dialogue optimization method based on multi-hop knowledge tracking, and the method comprises the following steps: accessing multi-source heterogeneous data, carrying out the preprocessing, and constructing a heterogeneous knowledge graph; dynamically updating the heterogeneous knowledge graph to generate a graph embedding representation set; the atlas embedding representation set is input into an improved heterogeneous graph Transform, and multi-hop path information is generated; in the multi-round semantic memory enhancement layer, multi-round memory fusion representation is generated; inputting the multi-round memory fusion representation into a type specific transformation layer, and outputting a reasoning result; jointly constructing a dialogue state vector by the reasoning result and current input and historical round information of the user, and generating candidate reasoning paths and structured knowledge responses; and outputting an explanatory result based on the candidate reasoning path and the structured knowledge response. According to the method, the improved heterogeneous graph Transform is adopted, and multi-hop knowledge tracking optimization of multiple rounds of dialogues is achieved.

The invention discloses a memory processing method and device, equipment, a storage medium and a program product. The method comprises the following steps: receiving a query request of a user, and identifying a query intention type of the query request; determining target weight configuration of a plurality of memory types based on the query intention type, wherein the plurality of memory types comprise plot memory, semantic memory, programmed memory and emotion memory; determining a semantic similarity score of the candidate memory data according to the semantic similarity and the target weight configuration, and determining a comprehensive score of the candidate memory data according to the semantic similarity score and the semantic strength score; and determining initial target memory from the candidate memory data according to the comprehensive score, determining associated memory data of the initial target memory based on a co-activation association relationship among the memory data, and taking the initial target memory and the associated memory data as target memory data.

With the development of IoT, intelligent sensing, and artificial intelligence technologies, various intelligent systems collect massive amounts of multimodal data daily, including images, videos, audio, tactile signals, and sensor readings. This data contains rich environmental information, user behavior, and event records, forming a crucial foundation for systems to understand the world and make decisions. However, existing technologies suffer from modal silos, fragmented information, lack of proactive utilization, and privacy risks. Therefore, a method and system are needed to unify the internalization of multimodal data into searchable and associative semantic memories, supporting proactive cross-modal utilization. This application provides a multimodal data internalization method, system, and computer-readable storage medium, aiming to enable various intelligent systems to uniformly transform multimodal data such as images, sounds, and touch into timestamped semantic descriptions, forming searchable and associative memory timelines, and supporting memory-based proactive behavior.

The application relates to the technical field of EEG emotion recognition, and provides a cross-subject EEG emotion recognition method and system based on decoupled hybrid meta-learning, which comprises the following steps: constructing a backbone network which is decoupled and connected in series by a cross-subject invariant feature encoding module and a sparse dynamic graph topology adaptation module; firstly, pre-training the cross-subject invariant feature encoding module based on a supervised contrast loss and an adversarial loss, fixing the parameters of the cross-subject invariant feature encoding module after training convergence, and outputting general node features; then, constructing a double-path hybrid meta-learning architecture: in the inner loop stage, a classification head is accessed for parameter adaptation, and then a scene semantic memory reasoning module is accessed for memory construction; in the outer loop stage, the prediction results of the parameter optimization path and the memory cognitive path are weightedly fused, and global meta-parameters are updated. Through the method, the accuracy and robustness of cross-subject EEG emotion recognition are improved.

The invention relates to a robot visual positioning method and system based on causal inference, and belongs to the field of visual positioning. The method comprises the following steps of: shooting continuous color images by a robot, and simultaneously acquiring instantaneous acceleration and angular velocity to obtain an original image sequence and an IMU original data stream; performing causal feature decoupling through the original image sequence, and outputting essential semantic feature vectors; modulating and reinforcing the essential semantic feature vector through meta-learning to obtain an optimized essential semantic feature vector; fusing and optimizing the essential semantic feature vector and the IMU original data stream, and outputting a time-pose track; based on the time-pose track, physical constraint verification is carried out through a physical constraint verification formula, and an optimal pose is output; and storing the optimal pose and the corresponding optimized essential semantic feature vector into a semantic memory layer, and constructing an environment semantic map. Visual positioning of the robot is achieved through the technologies of causal inference, meta learning, sensor fusion and the like.

The invention discloses a semantic enhancement method and device for low-illumination video image change detection, and the method comprises the steps: a time sequence semantic memory mechanism module carries out the modeling of cross-frame long-term semantic consistency based on a dual-temporal image through bidirectional time sequence interaction and context semantic enhancement, and outputs a feature representation with long-term semantic consistency; performing multi-scale processing on the feature representation of the long-term semantic consistency to obtain multi-scale fusion features, dynamically integrating the multi-scale fusion features by a progressive multi-scale semantic fusion network module through adaptive weighting, semantic guidance and low-illumination enhancement strategies, and outputting multi-scale fusion feature representation; and obtaining a final change detection result according to the multi-scale fusion feature representation. The device comprises a memory and a processor. Through the systematic semantic enhancement design, the noise interference is effectively relieved, the cross-frame semantic consistency is kept, and the small target detection capability under the low illumination condition is remarkably improved.

The invention discloses a construction waste language guidance segmentation system and method fused with invalid instruction recognition, and belongs to the technical field of artificial intelligence. According to the method, after cross-modal information is extracted and a semantic memory token is output, parallel segmentation and verification are executed in a second Swin Transform Stage, a third Swin Transform Stage and a fourth Swin Transform Stage of a Swin Transform visual coding network based on the semantic memory token and a negative memory token, candidate segmentation masks and existence probability scores are obtained, decision output is carried out through a gating mechanism with a preset confidence threshold, and the candidate segmentation masks and the existence probability scores are obtained. The output instruction is valid and invalid; and finally, after conditional logic modification is executed, joint loss is calculated, back propagation and optimization are carried out, and model training is carried out, so that the robust language guide segmentation system and method for invalid instruction interference are realized. The key problem that invalid instructions cannot be recognized and rejected during construction waste sorting is innovatively solved; and a powerful technical guarantee is provided for realizing real, reliable and efficient intelligent sorting of the construction waste.

The present disclosure relates to the technical field of text processing, and provides a text generation method and device, electronic equipment and a readable storage medium. The method comprises: performing memory fragment extraction processing on short-term memory data to obtain feature data corresponding to the short-term memory data, episodic memory data and semantic memory data; performing memory unit construction processing on the above data to obtain retrieval memory unit data and context memory unit data; performing similarity matching processing on the retrieval memory unit data based on query data to obtain target retrieval memory unit data; performing association processing on the target retrieval memory unit data and the context memory unit data to obtain target context memory unit data; and performing text generation processing on the query data and the target context memory unit data to obtain target dialogue response data. In this way, the efficiency and accuracy of retrieval matching are enhanced, the long-term reasoning and context understanding capabilities are enhanced, and the retrieval recall accuracy is improved.

The application provides a visual scene-oriented counterfactual verification memory construction method and system, which can obtain a target visual question and answer sample, including a target image and target question and answer information for the target image; input the target image and the target question and answer information into a visual intelligent agent for rule understanding, output candidate memory rules containing at least causal visual factors and non-causal interference factors; perform counterfactual image editing on the target image by changing one of the non-causal interference factors and the causal visual factors while keeping the other unchanged, respectively generate an interference edited image and a causal edited image, verify the interference edited image and the causal edited image, in the case of passing the verification, determine the candidate memory rules as target semantic memories, and determine the interference edited image as a target counterexample memory, and combine the target visual question and answer sample together to update a new memory item to a memory library of the visual intelligent agent, thereby improving the quality of the memory library.

This application belongs to the field of artificial intelligence and healthcare technology, specifically disclosing a multi-memory large model enhancement method and system based on adaptive resonance theory. The method includes: obtaining memory units from the plot memory module based on a new medical dialogue event; processing the new medical dialogue event through a large model based on multi-source memory; multi-source memory is constructed based on information stored in memory units, working memory modules, and semantic memory modules; the plot memory module is used to: determine candidate plot prototypes by analyzing attribute similarity; determine whether there is resonance between the new medical dialogue event and the candidate plot prototype; if there is resonance, use historical dialogue events in the event set of the resonant plot prototype as memory units and update the resonant plot prototype; if there is no resonance, create a new plot prototype. This application enables effective memory management and improves the stability and accuracy of large model reasoning in medical dialogue scenarios.

The invention belongs to the technical field of artificial intelligence and medical health, and particularly discloses a multi-memory large model enhancement method and system based on an adaptive resonance theory, and the method comprises the steps: obtaining a memory unit from a plot memory module based on a new medical dialogue event; processing a new medical dialogue event through a large model based on multi-source memory; the multi-source memory is constructed based on information stored by the memory unit, the working memory module and the semantic memory module; the plot memorizing module is used for determining candidate plot prototypes by analyzing attribute similarity; judging whether the new medical dialogue event and the candidate plot prototype resonate or not; if so, taking the historical dialogue events in the event set of the resonated plot prototype as a memory unit, and updating the resonated plot prototype; and if resonance does not exist, newly building a plot prototype. Through the method and the device, effective memory management can be realized, and the reasoning stability and accuracy of a large model in a medical dialogue scene can be improved.

The embodiment of the application discloses a memory management method for an LLM agent, an electronic device and a storage medium, wherein the memory management method for the LLM agent comprises the following steps: thinking / acting / observation generated by interaction is automatically divided and absorbed into structured work slots by the LLM in a review window; each work slot is filtered and routed according to a preset factor, and three types of long-term memory, namely semantic memory, episodic memory and procedural memory, are generated respectively; similarity threshold gating is adopted for updating or inserting long-term memory writing; and different retrieval strategies are adopted for the work slots and the long-term memory in the memory access stage.

The application discloses a memory optimization system for a large language model, comprising: a dynamic semantic memory modeling module, which generates context-aware memory vectors through a bidirectional Transformer and supports incremental updating to 100 million entity records; an adaptive memory resource scheduling module, which dynamically allocates memory pool partitions based on inference load, stores high-frequency memories in the working area, and enables quantization compression in the archive area; a memory decay and enhancement module, which adopts a double-factor decay mechanism, performs enhanced storage on high-value memories, and prunes them regularly; a cross-scenario memory migration module, which realizes cross-domain adaptation through a domain adapter and contrastive learning; and a memory conflict resolution module, which identifies conflicts based on a rule base and resolves contradictions using a multi-factor voting mechanism. The application improves memory retention integrity and retrieval efficiency, optimizes resource utilization, enhances cross-scenario adaptation capability, resolves memory conflicts, and improves model reliability.

The invention discloses a business experiment report automatic generation system and method based on deep learning, and the method comprises the following steps: obtaining an experiment data set, executing the structured processing, and generating an experiment fact input structure; constructing a hierarchical semantic reasoning graph structure, and establishing a cross-layer semantic association relationship; constructing an improved NBFNet reasoning model, executing cross-layer path relaxation propagation processing, and generating a candidate reasoning path set; constructing a path-level semantic memory structure and executing cumulative updating processing to generate a path semantic representation set; in the improved NBFNet reasoning model, multi-path competition type conclusion forming processing is executed, and a conclusion support distribution result is generated; executing conclusion type judgment processing to generate a conclusion type set; and carrying out commercial experiment report generation processing to generate a commercial experiment report. Structured reasoning and traceable expression of commercial experiment conclusion generation are realized, and report reliability and consistency are improved.

The invention provides a dynamic hierarchical semantic memory method. The dynamic hierarchical semantic memory method comprises a hierarchical encoder, a dynamic memory management and access mechanism and context-based semantic resolution and fusion. A multi-level and structured dynamic semantic memory system can be constructed, surface sequence information of a text can be captured at the same time, the method is used for local coherence and deep semantic architecture and global understanding, and the most relevant part can be dynamically activated according to a current task.