18 results about "Retrieval algorithm" patented technology

This application relates to the field of program repair processing technology, and in particular to an automatic program repair method and system based on a multi-level tree structure of a large language model. The method includes: using an information retrieval algorithm to retrieve a set of similar code examples corresponding to the program code to be repaired from a pre-defined code library; determining the model input content based on the set of similar code examples and the program code to be repaired; constructing a thought forest based on the model input content; performing error cause analysis based on the thought forest to obtain a set of high-confidence error causes; constructing a repair forest based on the set of high-confidence error causes; formulating repair strategies based on the repair forest to obtain a set of repair strategies; and performing autoregressive decoding operations based on the program code to be repaired, the set of similar code examples, and the set of repair strategies to obtain the patch code corresponding to the program code to be repaired. This application facilitates improvements in the accuracy and stability of the program repair process.

This invention discloses a method and device for auxiliary diagnosis of gas turbine faults, belonging to the field of industrial equipment fault diagnosis technology. Utilizing intelligent retrieval algorithms and efficient database support, it can quickly match relevant cases from a massive database of historical cases and provide diagnostic suggestions when a gas turbine experiences a warning or anomaly, shortening the fault diagnosis time. By characterizing historical fault cases of the gas turbine and combining them with current operating background parameters for multi-dimensional, dynamically weighted similarity matching, it can more accurately identify historical cases most similar to the current abnormal operating conditions, thereby improving the accuracy of fault diagnosis. In the specific application scenario of gas turbine power plants, it demonstrates unique technical contributions and significant practical value, which is of great significance for ensuring the safe and stable operation of gas turbines, improving power generation efficiency, and reducing operation and maintenance costs.

The application provides a retrieval enhancement generation method and device, and relates to the technical field of artificial intelligence. The retrieval enhancement generation method comprises the following steps: constructing a word element dictionary tree according to a pre-constructed knowledge graph; inputting a to-be-reasoned question and the word element dictionary tree into a pre-constructed relation type generation model to obtain a relation type set; performing entity recognition on the to-be-reasoned question, and mapping the recognized entity to an entity node of the knowledge graph to obtain a subject entity set; taking the subject entity set as a starting point, searching in the knowledge graph under the constraint of the relation type set by using a k-BET reasoning subgraph retrieval algorithm to obtain a target reasoning subgraph; converting a triple in the target reasoning subgraph into a preset format, and inputting the triple after the format conversion and the to-be-reasoned question into a pre-constructed reasoning result generation model to obtain a generation result, so that the retrieval enhancement generation with complete structure, faithful semantics, and consideration of efficiency and resource consumption is realized.

The application discloses an interactive self-service analysis and retrieval system based on a large model, which comprises a data source management module, a retrieval library construction module, an algorithm development module, a task scheduling management module and a service publishing module; the data source management module pre-processes initial information data to obtain structured information data; the retrieval library construction module constructs a retrieval library according to the structured information data; the algorithm development module extracts data features in the structured information data and recommends an adaptive retrieval algorithm according to the data features; the task scheduling management module performs retrieval analysis in the retrieval library according to the adaptive retrieval algorithm and an input retrieval instruction; and the service publishing module is used for converting the retrieval analysis result into an API service that can be called. The application aims to solve the pain points of the existing data analysis technology, such as fragmented functions, high operation threshold and lack of special text processing modules. The application provides an efficient solution for data modeling, text mining and service deployment.

This application provides a method, apparatus, device, and storage medium for self-correction of a medical visual language model based on a dynamic experience base (DEKB), relating to the field of medical visual language processing technology. The method includes: when the initial diagnostic result of the medical visual language model for a current clinical case does not match the fact label, constructing the current case as a structured experience unit and storing it in a dynamic experience knowledge base; upon receiving a query, retrieving historical experience cases related to the new query from the dynamic experience knowledge base; using the historical experience cases as contextual prompts to guide the medical visual language model in chain-like thinking, generating a corrected reasoning result. By constructing an endogenous dynamic experience knowledge base, designing a deep attribution analysis mechanism, and employing a dual-threshold retrieval algorithm, this application enables DEKB to transform the model's historical errors into structured knowledge that can be used for future reference, significantly improving the robustness and generalization ability of medical model image diagnosis reasoning.

This invention belongs to the field of computer information processing and proposes a document retrieval optimization method based on entity association. It performs secondary optimization while retaining the original retrieval algorithm, which not only improves the accuracy of the retrieval but also avoids the accuracy and efficiency problems caused by algorithm modifications. By performing entity recognition on both the document set and the user's search content, and calculating the association relationships between entities and documents, and entities and search content respectively, and comprehensively applying these association coefficients, it can uncover deep semantic relationships between search content and documents, and accurately locate documents that match the search intent. By calculating entity association degrees and re-ranking them according to these degrees, the accuracy of the search results is improved, resulting in a higher degree of alignment with user needs and allowing users to more easily obtain the information they require.

The application discloses a kind of prompt enhanced online social network privacy perception method, system, equipment and medium, method includes: depth collection analysis social network dataset, based on the dataset constructs the field knowledge graph containing multidimensional privacy disclosure mode, with hybrid retrieval algorithm retrieves privacy trigger word, entity and disclosure mode from graph with input text semantic highly relevant as result, result is integrated into dynamic prompt word template generation enhanced prompt, input large language model carries out fine-grained privacy identification.The application realizes the change from "entity recognition" to "entity-situation relationship determination" by association labeling strategy, combined with the situational knowledge injected by dynamic prompt word mechanism, so that the model can accurately distinguish between real privacy disclosure and public information description.In dealing with ambiguous expressions and new network language, due to the continuous updating ability of knowledge graph and the generalization performance of contrast learning feature representation, the recognition accuracy is significantly improved compared with existing methods.

The application discloses a CSV large file fast retrieval method and device, equipment and storage medium, and relates to the technical field of computers, which comprises the following steps: determining a first address, a second address and an end address from a CSV file stream based on a retrieval data request; wherein the first address is a title address of the CSV file stream, the second address is a starting address of the first data in the CSV file stream in the memory, and the end address is an end address of the last data in the CSV file stream in the memory; taking the number of bytes between the first address and the second address as a target step; determining a center address of the CSV file stream according to the second address and the end address; and starting from the center address, traversing the CSV file stream between the to-be-traversed addresses in the forward direction of the center address and the backward direction of the center address by using the target step, until the retrieval data request is met. The retrieval algorithm is optimized, and the effect of fast retrieval and low consumption in the CSV large file is achieved.

This invention relates to the field of knowledge base technology, specifically to a knowledge base retrieval method supporting compression. It includes the following steps: S1, establishing a B+ tree as the index structure of the knowledge base, and compressing the B+ tree index structure using prefix compression technology; S2, compressing the data in the knowledge base using an intelligent differential compression algorithm; S3, adjusting the interaction between the index structure and compressed data using an intelligent index retrieval algorithm based on the B+ tree index structure; S4, monitoring performance indicators in real time and periodically checking the integrity of the data. This knowledge base retrieval method supporting compression optimizes data storage through an intelligent differential compression algorithm, significantly reducing storage space, and employs an intelligent index retrieval algorithm based on a B+ tree to achieve intelligent and fast data retrieval, optimize query paths, reduce decompression time, and improve retrieval efficiency.

The application provides a car fault knowledge sharing method and system, and belongs to the technical field of car maintenance. The method comprises a knowledge base construction and updating process and a retrieval process based on the knowledge base. In the knowledge base construction stage, a system administrator defines multiple car professional fields and assigns user permissions, maintains user submission of fault knowledge update requests, and updates the knowledge base after approval by the user for auditing. In the retrieval stage, the terminal user inputs fault description information, the system pre-processes the query information through a semantic retrieval algorithm, vectorizes the representation, calculates the cosine similarity with the fault nodes in the knowledge base, and generates retrieval results in descending order of matching degree. The application realizes the systematic accumulation and dynamic updating of car fault knowledge, provides an efficient and accurate knowledge retrieval channel, constructs a complete knowledge management closed loop, and significantly improves the car maintenance efficiency and fault diagnosis accuracy.

This invention provides a method and system for retrieving patient digital models based on event graphs. The method includes: designing a timeline-based event extraction and structuring method for unstructured or semi-structured data in electronic medical records, constructing an event graph of the electronic medical records; extracting key information from the medical record text in the event graph using an intelligent model; designing a patient digital model retrieval algorithm; establishing a key information retrieval and matching mechanism; and quickly searching for digital models similar to the current patient's condition from a historical case database, providing data support for subsequent diagnosis and treatment analysis. This invention can extract key information from massive heterogeneous clinical data, construct patient digital models with high clinical relevance, interpretability, and dynamic adaptability. Based on event graph technology and large-scale model applications, it improves the modeling accuracy and clinical applicability of patient data in the diagnosis and treatment of complex diseases, significantly improving the accuracy, completeness, and clinical relevance of patient digital models.

The application discloses a method, device and storage medium for retrieving data, and belongs to the field of communication. The method comprises the following steps: receiving a retrieval request, wherein the retrieval request comprises n first characteristic vectors, the n first characteristic vectors are characteristic vectors of first unstructured data to be retrieved, n is an integer greater than or equal to 1; obtaining a first retrieval algorithm based on n, wherein the first retrieval algorithm is a retrieval algorithm whose performance for retrieving data by using the n first characteristic vectors meets a first condition; retrieving at least one second unstructured data in a first corresponding relationship by using the first retrieval algorithm based on the n first characteristic vectors, wherein the first corresponding relationship is used for saving the corresponding relationship between the second unstructured data and second characteristic vectors of the second unstructured data. The application can improve the efficiency of retrieving data.

The application belongs to the technical field of natural language processing and information retrieval, and relates to an intelligent question and answer method based on structured semantic index and double-layer memory enhancement, which comprises six steps of query preprocessing, Agent-based multi-tool dynamic routing, structured and semantic enhanced index construction, adaptive context assembly, double-layer memory management, data closed loop and self evolution. By adopting a differentiated segmentation strategy for text, code and multi-modal data, combining query optimization, intelligent tool routing and hybrid retrieval algorithm, the technical defects of existing RAG systems, such as semantic breakage, low retrieval accuracy, lack of long-term memory and self-optimization, are solved. The application improves the code retrieval accuracy and context utilization rate, realizes personalized long-term service and automatic operation and maintenance, and is suitable for technical research and development, code library maintenance and intelligent question and answer scenes.

The application provides a specific paper information extraction method based on hybrid retrieval, comprising: obtaining a plurality of paper data of a specific technical field to be information extracted, and extracting text data and table data in the paper data; using a recursive blocking algorithm in a large language model programming framework to divide the text data into text blocks while preserving context; converting each text block into a vector representation using a pre-trained language model; classifying the indicators to be extracted, retrieving the text blocks and their vector representations through a multi-round hybrid retrieval algorithm combining keyword and semantic search, and obtaining the indicator categories to which each text block belongs; using a large language model to extract the indicators of the vector representation according to the indicator categories of the vector representation, and structuring the indicators into a standard format as the information extraction result of the paper data.

This invention discloses a multimodal agent-enhanced retrieval algorithm, belonging to the field of enhanced retrieval algorithm technology. The invention includes a multimodal text conversion unit, a vector database construction unit, and a multimodal knowledge graph (MMKG) construction unit. It enhances cross-modal reasoning capabilities, solving the modality disconnect problem in existing technologies: establishing visual-text associations through MMKG. It improves the efficiency of multi-source knowledge synthesis by employing plug-in parallel retrieval. Through consistency verification of the decision-making agent and expert refinement, it reduces the answer illusion rate, i.e., the proportion that conflicts with facts; MMKG embeds visual data storage paths, and web retrieval provides URL tracing, allowing users to trace the source of each fact in the answer, thus solving the black-box reasoning problem in existing RAGs.

This application discloses an intelligent safety monitoring method, device, equipment, and storage medium, relating to the field of artificial intelligence technology. The method includes: when a pre-defined intelligent agent receives abnormal data from the field, it uses retrieval enhancement generation technology to retrieve target retrieval text related to the abnormal data from production safety documents; it uses a pre-defined large language model, a pre-defined multimodal model, and a heuristic retrieval algorithm to perform task planning on the target retrieval text to obtain target tasks and corresponding execution actions, storing the target tasks in a task list; it executes the target tasks in the task list using a pre-defined execution strategy and execution actions, and judges the deviation between the target task execution result and the expected task execution result; if the deviation between the target task execution result and the expected task execution result is greater than a pre-defined deviation threshold, it generates a feedback report corresponding to the abnormal data, so as to perform anomaly handling based on the feedback report. This comprehensive judgment of abnormal data avoids false alarms.

The application provides an intelligent operation and maintenance root cause analysis method and system based on time sequence topology and streaming memory, comprising extracting a hard topology skeleton from multi-source operation and maintenance data, anchoring soft semantic information to the skeleton to construct a mixed heterogeneous graph; dynamically storing and updating the graph index through a streaming double-layer memory architecture containing short-term and long-term memory layers; in response to a fault trigger signal, cooperatively retrieving a root cause evidence subgraph from the double-layer memory based on a time sequence causal weighting retrieval algorithm integrating a time decay factor and a reverse causal walk; and generating a root cause analysis result based on the subgraph. The method eliminates the logical illusion of root cause analysis, solves the problem of knowledge update lag, realizes accurate and real-time positioning of complex system cascade fault root causes, and adapts to the high dynamic operation and maintenance needs of power information systems.

The application discloses a kind of intellectual property big data intelligent retrieval method and system based on semantic understanding, it is related to big data retrieval technical field, first, collect intellectual property original data, store into cloud original distributed database after uniform format etc.Processing;Then, construct quality annotation dataset and semantic verification model to generate quality label and store into cloud library;Again, parse user search request to generate semantic vector;After hierarchical retrieval matching, low-quality data is corrected and matching score is calculated to generate initial result;Finally, collect user behavior data to update algorithm parameters, re-verify and write back quality information, the present application builds cloud original distributed database and stores intellectual property original data and standardizes processing, constructs quality annotation dataset and semantic verification model to realize quality accurate determination;At the same time, the field analysis of search request is resolved, hierarchical retrieval matching is carried out, the retrieval algorithm is corrected and optimized through closed loop system, the accuracy and efficiency are improved, and user demand is adapted.