59results about How to "Improve retrieval accuracy" patented technology

The invention belongs to the technical field of search engines, and particularly relates to an optimized search engine construction method and system based on feature deep learning, and the system comprises a search operation unit, a monitoring and early warning unit and a management and control terminal, the search operation unit comprises a multi-dimensional semantic feature mining module, a user implicit intention analysis module, a deep feature fusion modeling module and a dynamic retrieval strategy generation module; a multi-dimensional semantic feature set and a user intention feature set are formed through a multi-dimensional semantic feature mining module and a user implicit intention analysis module, and a deep feature fusion modeling module performs deep fusion of data features and user intention features through an attention mechanism and cross validation optimization. The dynamic retrieval strategy generation module is used for providing high-quality feature support for retrieval strategy generation, the accuracy of retrieval results is guaranteed, the dynamic retrieval strategy generation module generates dynamic strategies adaptive to different requirements and data states on the basis of reinforcement learning, and the retrieval accuracy and retrieval efficiency of a search engine are remarkably improved.

The invention relates to the technical field of digital archive management and information retrieval, and discloses a digital archive multi-modal data semantic enhancement fusion retrieval method and system.The method comprises the steps that a policy cycle time axis and a policy term evolution graph are constructed, tense logical reasoning is conducted on archive seals, and permission effectiveness evolution is derived; the temporal permission feature vector and the content semantic vector are fused to generate a multi-modal representation vector, and cross-policy-cycle semantic enhancement retrieval is realized by combining query expansion and temporal permission filtering, so that the problems of missing detection and misjudgment of policy and regulation archives in seal permission historical evolution and term cross-cycle retrieval are solved.

The embodiment of the invention provides a multi-modal data retrieval method and device, equipment, a medium and a product, and the method comprises the steps: obtaining a natural language problem of a user, carrying out the intention understanding of the natural language problem, and obtaining the query intention of the natural language problem; performing retrieval path selection on the query intention, and determining at least one target modal type matched with the natural language question; the natural language question is distributed to vector databases corresponding to the target modal type to be retrieved, and candidate multi-modal retrieval results of the natural language question are obtained; and verifying the candidate multi-modal retrieval result according to a pre-constructed multi-modal business knowledge graph to obtain a target multi-modal retrieval result of the natural language problem. By utilizing the method, a collaborative mechanism of semantic comprehension, routing decision and knowledge enhancement is introduced, so that the cross-modal semantic comprehension capability and the information retrieval performance are remarkably improved.

The invention relates to the technical field of information retrieval, in particular to a large-scale semantic retrieval optimization method based on deep learning, which comprises the following steps: acquiring to-be-retrieved text data, and converting the to-be-retrieved text data into semantic vector representation through a deep learning model; carrying out topological structure analysis on the semantic vector representation to obtain topological characteristic information of a semantic space; on the basis of the topological characteristic information, a topological perception hierarchical navigable small-world graph index with a multi-layer structure is constructed, and each layer comprises a node set and a connection relation thereof; dynamically optimizing and adjusting the topology-aware hierarchical navigable small-world graph index according to an actual query mode and system performance feedback; and receiving a query request, converting the query request into a query vector, executing multi-layer navigation search in the topology-aware hierarchical navigable small-world graph index, and returning a retrieval result related to the semantics of the query vector.

The invention discloses a video content retrieval method and device based on television system voice interaction and a terminal, and relates to the technical field of video processing, and the method comprises the steps: when a video is played for the first time, extracting a picture frame from the video at a preset frequency, converting the picture frame into a multi-dimensional image feature vector and a corresponding video timestamp, and carrying out hierarchical storage in a database, constructing vectorized data containing visual semantic information; obtaining a voice retrieval instruction, performing intention recognition and semantic understanding, extracting a detection keyword, and generating a multi-dimensional retrieval feature vector; calculating a matching degree between the multi-dimensional retrieval feature vector and a multi-dimensional image feature vector of a video picture frame stored in a database, and screening out picture frames of which the similarity is higher than a preset similarity threshold to form a retrieval candidate matching set; and determining matched picture playing. The video content retrieval method is efficient, accurate and high in interactivity, and retrieval experience and operation efficiency of the user in the video watching process are remarkably improved.

The invention discloses a method for constructing, updating and retrieving an action memory library, which belongs to the technical field of computer vision and comprises the following steps of: constructing a training data source with time sequence diversity; initializing an action memory library containing a plurality of learnable prototype matrixes and a double-flow interaction network; memory bank evolution is executed, an action prototype is retrieved by utilizing a query stream, a current memory state is dynamically generated by combining a memory state updating gate mechanism with a historical state, and dynamic memory is injected into a feature space by utilizing memory driving graph convolution; synchronously updating parameters based on multi-target loss, and driving a memory bank to evolve into optimal structured prior; and finally, performing structured reasoning on the to-be-detected sequence by using the optimal memory bank. According to the method, structured priori is constructed by mining a spatio-temporal topology mode of a human body action sequence, and hierarchical memory evolution and double-flow depth interaction are combined, so that the problem of depth ambiguity in a monocular vision task is effectively solved, geometric structure distortion is corrected, and the accuracy of action posture estimation is remarkably improved.

The video character retrieval method, device, equipment and storage medium of the present application, wherein the method comprises: establishing an original video database, obtaining a target video from the original video database; extracting a character segment in the target video; extracting a character audio in the character segment; performing semantic analysis on the character audio to obtain an audio text; combining the audio text and a video text in the target video to obtain a combined text; and using a preset character identity recognition model to extract a character identity in the combined text, wherein the preset character identity recognition model is obtained by training a convolutional neural network. The combined text combining the audio text and the video text includes the character name and the character relationship of the target video, and the character identity in the combined text can be automatically extracted by the trained preset character identity recognition model, thereby achieving high retrieval accuracy and efficiency.

The invention provides a software V-shaped development process oriented RAG retrieval method and system, and the method comprises the steps: receiving a technical document, and carrying out the segmentation and recognition of the document, so as to obtain a document fragment set; extracting an engineering entity, an identifier and an attribute from the document, and constructing a binding relationship between a document fragment and the entity; constructing a knowledge graph containing nodes and node association relationships in a graph database or a triple; constructing a multi-dimensional vector representation of the document fragment, and writing a vector into a vector database; performing semantic retrieval on matched candidate document fragments based on a user query request, and performing neighborhood expansion on the target entity in the knowledge graph to obtain a candidate graph; aligning the candidate document fragments with the candidate atlas, and selecting a preset number of fragments which are ranked in the top to be aggregated into an evidence packet; and inputting the evidence packet and the generation task instruction into a large language model, and generating a retrieval result based on a predetermined constraint decoding strategy. By means of the scheme, the consistency of all stages in RAG retrieval can be guaranteed, and the content retrieval precision is improved.

This application discloses a trademark retrieval database generation, retrieval method, and apparatus that enhances local features. The method involves constructing a multi-classification training set for trademark graphic elements and training it in an image multi-classification model to obtain a trademark element multi-classification model; constructing a target detection training set for important graphic elements based on trademark data and training it in an image target detection model to obtain a target detection model; inputting all trademark data into the trademark element multi-classification model for prediction and selecting trademark elements as field information; inputting all trademark data into the target detection model to detect important graphic element targets and extracting them into sub-images; inputting all sub-images and the original trademark data image into a trademark encoding model for encoding to obtain encoding vectors; and inputting the field information and encoding vectors into a vector database to obtain the trademark retrieval database. Using the trademark retrieval database provided in this application can significantly improve trademark retrieval results.

The invention discloses an educational academic literature tracing method and system based on retrieval enhancement generation. The method comprises the following steps: performing fine-grained analysis on the multi-source literature, constructing a document object containing fields such as a title, an abstract, a methodology, an experimental result and a conclusion, and establishing a field-level index and a citation network feature score; identifying an academic intention queried by a user by utilizing a large model, dynamically loading a weight mapping table according to the academic intention, and performing weighted rearrangement on multiple paths of retrieval results to obtain candidate core literatures; generating a traceability text with an explicit label based on the candidate literature; and executing reference consistency verification by using the natural language inference model, and executing illusion correction according to a verification result. According to the method, through structured field analysis and dynamic rearrangement of intention driving, the problems of semantic fragmentation and reference fictition when a general RAG is used for processing complex academic literatures are solved, and the academic traceability preciseness and accuracy are remarkably improved.

The application provides a retrieval method, device and equipment for software product prototype information, a medium and a product. The method comprises the following steps: obtaining a plurality of prototypes of a target software product and corresponding associated information of each prototype, wherein the associated information comprises prototype source files, software product information and plug-in resources used by the prototypes, and the software product information comprises function characteristic information and version information. For any prototype, the prototype source files, the software product information and the plug-in resources used by the prototype are associated and bound to generate an associated data set, and a prototype library containing a function characteristic dimension and a version dimension is constructed based on the associated data set of each prototype. Then, in response to a retrieval condition input by a user, the prototype library is retrieved according to the function characteristic dimension or the version dimension to obtain matched prototype information. The application improves the prototype retrieval efficiency and accuracy.

The application provides a medical question and answer method and system based on cooperative dual-source retrieval enhancement generation, which comprises the following steps: according to an input original medical question and candidate items, performing task self-adaptive query rewriting through a large language model to generate differential diagnosis keywords, entity enhanced queries and hypothetical medical abstracts; for the differential diagnosis keywords, performing an iterative network retrieval with a reflection mechanism to obtain network candidate evidence; for the entity enhanced queries and the hypothetical medical abstracts, performing a hybrid retrieval in a local medical knowledge base to obtain local candidate evidence; aggregating and deduplicating the network candidate evidence and the local candidate evidence, obtaining the aggregated heterogeneous evidence, performing deep semantic correlation scoring through a cross-encoder model, screening out a target evidence set, and splicing the original medical question as context input into a generative large language model to generate a final medical answer. The application improves the accuracy, timeliness and explainability of the medical question and answer system.

This application provides a method for loading dynamic capability configurations for a home-based embodied intelligent robot, comprising: Step 1: In response to a new hardware access signal, performing local hardware pairing and service registration; after the pairing and service registration are successful, retrieving a dynamic capability extension package corresponding to the accessed hardware, containing capability execution logic and interface configuration requirements, from a preset capability repository, and generating a global capability mapping table recording component installation status and hardware / software interface occupancy information based on the dynamic capability extension package; Step 2: In response to a received task instruction, performing capability retrieval in the global capability mapping table to determine the hardware readiness status and installation status corresponding to the target logic component required for the task to be executed, and then performing on-demand installation and data loading processing on hardware that meets preset criteria to produce a set of loaded capabilities containing the running entity of the target logic component.

This invention relates to the field of video data processing and multimedia information retrieval technology, specifically a video big data retrieval method and system based on semantic tags. The method includes: a data mapping step: mapping a video stream into a sequence of feature vectors using a multimodal encoder; a difference calculation step: calculating the rate of change of differences between adjacent feature vectors to generate a semantic velocity vector; an anchor point extraction step: extracting semantic abrupt change event anchor points based on a comparison of the velocity vector magnitude with a threshold; an index construction step: constructing a differential manifold index with semantic transitions as nodes and time spans as edges; a dynamic dimensionality reduction step: constructing a dynamic mask matrix based on the retrieval request, projecting nodes to a low-dimensional subspace; and a matching output step: calculating the matching degree in the subspace and outputting the result. This invention significantly improves retrieval speed and signal-to-noise ratio, and substantially reduces storage volume through index event anchor points and dynamic subspace folding.

The invention discloses a multi-model paper retrieval method oriented to academic questions and answers, which relates to the technical field of natural language processing and information retrieval, and comprises the following steps: firstly, constructing a unified corpus and a training data set, and respectively encoding by using models in a first model set and a second target model to generate a document vector set; screening difficult negative samples based on the initial retrieval result of the second model, constructing a comparative learning sample, performing fine adjustment on the comparative learning sample, and recoding a corpus; and taking the fine-tuned second target model and the models in the first model set as a model group, respectively performing target query and performing similarity retrieval by adopting each model in the model group to obtain a corresponding original similarity matrix, screening documents based on the original similarity matrix, and generating a first target document list of the target query. According to the method, the accuracy and robustness of academic literature retrieval can be effectively improved.

The application provides an agent dynamic routing and multi-tool collaborative question and answer method for standard class text, and relates to the technical field of natural language processing. The application realizes intention recognition and dynamic routing through an agent, collaboratively schedules three types of special retrieval tools, accurately adapts the composite intention of standard query, breaks through the limitation of a single retrieval path, and significantly improves the accuracy of standard text retrieval and response quality. At the same time, a strict calling closed loop is constructed to evaluate the integrity of the retrieval result, and secondary routing is triggered when the information is insufficient; when the question is ambiguous, the candidate standard number is forced to be bounded and the knowledge base is deeply mined, the path of fabricating answers is blocked from the architecture level, and the output content is ensured to be traceable and compliant. The method relies on multi-tool collaboration and closed loop scheduling to provide reliable and real information support for professional question and answer of standard text, guarantee high robustness and rigor, and adapt to the needs of strict professional scenarios.

This invention relates to the field of intelligent fabric matching and recognition technology, and discloses a fabric matching method and system based on multimodal information matching. The method includes: semantically structuring and enhancing user input text; acquiring the original fabric image and performing mask extraction, texture enhancement, and normalization processing; calculating the matching degree based on a cross-modal semantic alignment mechanism, and outputting fabric recommendation results by combining multi-factor joint ranking. Compared to existing technologies that use simple text retrieval or single-channel feature matching, especially when user descriptions are non-standard, fabric image textures are complex, or structures are repetitive, failing to achieve accurate fabric location and consistent semantic matching between text and images, this application improves fabric retrieval accuracy in open descriptive contexts by introducing an industry ontology-driven semantic enhancement mechanism and a saliency-guided cross-modal matching architecture, achieving multi-granular modeling and fine-grained semantic alignment of fabric semantic features.

The invention discloses a multi-modal heterogeneous data increment embedding fusion method, equipment and a medium, mainly relates to the technical field of multi-modal heterogeneous data, and aims to solve the problems of difficult neighborhood structure maintenance, insufficient cross-modal information fusion, single and fixed index structure and the like in the existing scheme. Comprising the following steps: converting modal data in a multi-modal heterogeneous data set into hidden space representation, and fusing the hidden space representation by utilizing a cross-modal attention mechanism to obtain historical embedding representation of a preset d dimension; when the newly-added multi-mode heterogeneous data set appears, calculating a newly-added embedding representation of a preset d dimension; according to the newly-added multi-modal heterogeneous data set, the newly-added embedding representation, the historical multi-modal heterogeneous data set and the historical embedding representation, constructing an adjacency relation graph of the newly-added multi-modal heterogeneous data set; and inputting the adjacency relation graph into a pre-trained incremental graph neural network to obtain an embedded representation fusing the adjacency relation.

The invention relates to the technical field of big data management and retrieval enhancement generation, and particularly discloses a college data management large model platform based on a vector database and retrieval enhancement, comprising a multi-modal vectorization engine integrating a multi-modal large language model to realize deep semantic coding and representation learning of education data; the intelligent vector retrieval system realizes high-precision rapid retrieval based on neural information retrieval and learnable indexes; the advanced RAG inference engine improves the generation quality through an adaptive retrieval strategy, a multi-hop inference mechanism and knowledge fusion; the intelligent educational knowledge engine provides domain knowledge support and intelligent analysis based on dynamic knowledge graph and educational ontology modeling. By integrating the latest artificial intelligence technology, an advanced college data management platform is constructed, powerful technical support is provided for education informatization, and important theoretical significance and practical value are achieved.

The application discloses a tree document retrieval method, device, equipment and storage medium. The method comprises the following steps: obtaining a pre-stored document tree, the leaf nodes of the document tree correspond to document content, and the non-leaf nodes correspond to directory information; initializing a task tree representing a retrieval path, the initial state of the task tree comprises a root node, and the four-tuple information of the root node comprises an original question, a document tree root node path, document tree root node content and an empty history record list; inputting the information on the current task tree into a large language model, the large language model plans a retrieval path based on the task tree, and outputs text information returned by the root node of the task tree; and obtaining a retrieval result based on the text information returned by the root node of the task tree. The application proposes a document retrieval method based on a task tree and a document tree, uses the reasoning and multi-step thinking ability of the large language model, so that the model can further explore and reason according to the existing node information, and generate more accurate and comprehensive answers.

The invention provides a memory method for supporting hybrid retrieval and dynamic update in large law model application. The memory method comprises the following steps: acquiring a question input by a user; performing mixed retrieval according to questions of the user, and obtaining related memory information through secondary retrieval of law article fields and mixed retrieval of a conventional memory library; performing memory evolution according to the current user input and the mixed retrieval result, judging whether the user input is matched with the retrieval result, updating the existing memory or creating new memory according to the matching result, and updating the link relationship between the memories; and filtering and screening the mixed retrieval result to obtain a final memory retrieval result. According to the method, the memory system can be adaptively optimized according to the actual use condition, the method has the advantages of combining multiple retrieval methods, the dual requirements of the legal field for accuracy and semantic understanding can be met at the same time, multiple application scenes such as legal consultation, case analysis and legal education are effectively supported, and then the efficiency and accuracy of legal services are improved.

This invention relates to the field of NLP technology and discloses a plug-in function testing method for a pre-trained power system model based on NLP. The method includes: receiving a technical supervision query question and performing vector encoding and retrieval to obtain an initial retrieval result set; obtaining reference text with standard citation annotations through weighted sorting and concatenating it with multi-level prompt word templates to obtain domain-structured prompt words; performing a bundle search using a large technical supervision domain model fine-tuned with LoRA low-rank adaptation to generate candidate answer texts; and conducting quality assessments including citation integrity detection, logical consistency detection, and entity illusion detection to obtain a qualified technical supervision standard answer. This method solves the technical problems of low standard retrieval accuracy, poor answer accuracy, lack of citation tracing, and insufficient domain customization capabilities in existing technologies, providing a complete technical solution for intelligent applications in the field of power technical supervision.

The application discloses a kind of academic question and answer-oriented multi-model paper retrieval method, it is related to natural language processing and information retrieval technical field, including: first, construct unified corpus and training dataset, utilize the model in first model set and second target model respectively encode generation document vector set;Then, based on the initial retrieval result of second model, difficult negative sample is filtered, and the contrast learning sample pair is constructed to fine-tune and re-encode corpus;With the model group of the second target model after fine-tuning and the model in first model set, each model in model group is executed similarity retrieval in parallel respectively, and the corresponding original similarity matrix is obtained, based on the original similarity matrix, the document is filtered, and the first target document list of target query is generated.The application can effectively improve the accuracy and robustness of academic literature retrieval.

The invention discloses a multi-step reasoning information acquisition method, system and device and a storage medium, which are corresponding schemes: performing semantic analysis and dynamic correction on task query by using a large language model to realize multi-step reasoning and adaptive optimization of query semantics; moreover, an incremental multi-path recall system fused by heterogeneous associated signals is designed, and keyword, semantic and structured features are fully fused through a multi-path, multi-view and multi-round progressive recall joint mechanism, so that high-coverage and high-precision candidate information recall is realized; in addition, a fine ranking scheme based on deep thinking (thinking chain) and reasoning expansion is provided, and interpretable logic analysis and accurate ranking are carried out on candidate information. Generally speaking, the information understanding, reasoning depth and retrieval accuracy of the model under the complex incidence relation can be remarkably improved while the calculation efficiency is kept, and universal technical support is provided for a high-reliability knowledge acquisition and scientific intelligent system.

This invention discloses a retrieval enhancement generation method and system based on contradiction detection feedback, aiming to solve the problems of single data source, insufficient retrieval accuracy, and lack of factual verification in existing technologies. The method mainly includes three core steps: First, document block storage, using OCR and VLLM technologies to extract features from text, tables, and images and uniformly vectorize and store them; second, multi-source query retrieval, using an iterative retrieval controller to introduce a self-reflective closed loop, dynamically evaluating and optimizing the quality of the retrieved content; and finally, contradiction detection generation, comparing the generated content with the source documents for factual accuracy, and feeding back any conflicts to the model for forced correction. This invention effectively expands the sources of multimodal knowledge, suppresses model "illusions," and significantly improves the accuracy and fidelity of the generated answers.

The embodiment of the invention provides a long text retrieval method and device based on a large model, equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the steps of obtaining a long text sample; executing an answer generation task on the key information sample and the question sample based on a preset large model to obtain an answer prediction sample of the question sample; performing task contribution evaluation on the attention head according to the key information sample and the answer prediction sample to obtain a task contribution score; wherein the task contribution score is used for representing the contribution degree of the attention head in the answer generation task; screening out a target retrieval head from the at least two attention heads according to the task contribution score; performing model updating on the preset large model according to the target retrieval head to obtain a target large model; and obtaining a target long text, and performing text retrieval on the target long text according to the target large model. The method can be applied to business systems needing a large number, such as financial science and technology and health medical treatment, and the accuracy of long text retrieval is improved.

The application discloses a dangerous chemical substance intelligent question and answer method and system based on CAS structured retrieval, which comprises the following steps: obtaining question data; performing semantic analysis on the question data to generate a structured SQL query instruction; constructing a hierarchical database, performing CAS hierarchical retrieval by using the structured SQL query instruction, and recalling authoritative information of dangerous chemical substances; constructing a constraint prompt word, jointly coding the user question, the authoritative information and the scene by using a large language model, and generating a constrained dangerous chemical substance management and emergency suggestion; and through a semantic analysis mechanism of "keyword extraction, question classification and SQL classification construction", combining a hierarchical retrieval strategy taking CAS number as an anchor point and a large language model assisted normalization CAS query mechanism, stable mapping of diversified names of dangerous chemical substances to unique CAS numbers is realized, retrieval difficulties and generation inaccuracies caused by diversified names of dangerous chemical substances and fuzzy questions are effectively relieved, and the accuracy, stability and traceability of information retrieval and scenario-based suggestion generation of dangerous chemical substances are improved.

This invention proposes a wearable device interaction method and apparatus, and a computer device based on memory information. The method includes: collecting multimodal raw data through a wearable device; performing event recognition and semantic extraction on the raw data of each modality, and outputting a single-modal semantic description; comprehensively analyzing all single-modal semantic descriptions within the same time period to generate an event chain and event context description; saving an event table, an association information table, and an index table to a database, wherein the event table stores the event chain and event context description, the association information table stores the relationships between events, and the index table establishes a data retrieval channel; automatically generating reminder tasks based on the data in the event table, association information table, and index table, and pushing them to the user when the task triggering conditions are met, wherein the task triggering conditions include at least one of time, location, scenario, and unfinished tasks. This invention achieves efficient and accurate active and passive intelligent interaction, improving user experience and system efficiency.

The application relates to a text retrieval method and device, computer equipment, a storage medium and a computer program product, relates to the technical field of computers, and can also be used in the field of financial technology or other related fields. The method comprises the following steps: acquiring a text to be retrieved; performing semantic feature extraction processing on the text to be retrieved to obtain initial semantic features of the text to be retrieved; performing reconstruction processing on the initial semantic features to obtain target semantic features of the text to be retrieved; the completeness of the semantics represented by the target semantic features is greater than the completeness of the semantics represented by the initial semantic features; and based on the target semantic features, target text matching the text to be retrieved is determined from a plurality of candidate texts. The method can improve the retrieval accuracy of text retrieval.