73 results about "Document summary" patented technology

A computer-implemented, machine learning method for generating explainable text summaries includes extracting a subset of sentences from an input document as an extractive summary and adding context to the extracted sentences to generate a prompt. A fluent summary is generated by using the prompt as input to a generative language model. Source information for a sentence from the fluent summary is determined by mapping the sentence from the fluent summary to a sentence in the extractive summary and the sentence from the extractive summary to a sentence from the input document. A transparent summary view is generated showing the sentence from the fluent summary along with the source information from the extractive summary and the input document for display on a user interface. The method has applications including, but not limited to medical AI, public safety and other machine learning applications for reliable and explainable document summarization.

The invention discloses a knowledge retrieval method based on a multilayer index, which comprises the following steps of: obtaining a knowledge base of corresponding classification according to a knowledge retrieval statement in combination with a knowledge classification label; obtaining a plurality of retrieval target documents through vector similarity analysis according to the knowledge retrieval statements in combination with the document abstracts in the knowledge base of the corresponding classification; and obtaining a plurality of similar text blocks through vector similarity analysis according to the knowledge retrieval statement in combination with the text blocks of the plurality of retrieval target documents, and outputting the similar text blocks as knowledge retrieval. According to the method, a three-layer index structure containing knowledge classification tags, document abstracts and text blocks is constructed, the classification tags are dynamically matched through a large language model to reduce the retrieval range, high-correlation documents are rapidly positioned in combination with document abstract vector similarity analysis, and context logic relations are reserved based on text block hierarchical vector matching. The problems of fuzzy classification, low efficiency of document recall and context breakage in the traditional technology are solved, and accurate retrieval from macroscopic to microscopic is realized.

The invention discloses a rule and regulation intelligent question-answering method and device based on two-stage retrieval and a large model, and belongs to the field of rule and regulation file-based intelligent question-answering, and the method comprises the following steps: carrying out file abstract extraction on each file based on a large model question-answering form, and constructing an abstract vector library; cutting and partitioning all the files, respectively vectorizing text boxes of all text blocks by adopting a sparse vector model and a dense vector model, and constructing a content vector library; in the first stage, files most relevant to questions of a user are retrieved, two vector libraries are retrieved in a mixed retrieval mode, and weighted calculation is conducted on retrieval on the two sides to obtain the first K most relevant files; in the second stage, the text blocks most relevant to the questions of the user are retrieved, and the first k most relevant text blocks are obtained through mixed retrieval; and during answering, taking the first k most relevant text blocks as reference texts, and enabling the large model to answer based on the reference texts. According to the invention, more accurate answers can be generated.

The invention provides a CAD file summary information automatic extraction method and device, a medium and a product, and relates to the technical field of computer picture word process.The method comprises the steps that a target CAD file is analyzed through a CAD file analysis development tool, and first word information and geometric attribute information corresponding to the first word information are obtained; retrieving the first character information according to the retrieval keyword to obtain second character information; according to the corresponding geometric attribute information in the second text information, carrying out positioning adjustment display through a CAD file analysis development tool to obtain a picture file; calling a first large language model to perform character recognition and arrangement processing on the picture to form a text list; and calling the second large language model to obtain summary information of the target CAD file according to a preset prompt template and the text list. According to the scheme, when a large batch of CAD files are processed, the processing efficiency is remarkably improved.

The invention relates to the technical field of medical care informatics, and discloses a medical questionnaire open question answer coding method, system and device and a medium, and the method comprises the steps: converting an answer document of a medical questionnaire open question into an answer document vector through a document embedding model; an aggregation hierarchical clustering algorithm is utilized to quickly and accurately identify answer document clusters with hierarchical relationships and different answer document themes in answer documents, so that complex information is structured, subsequent analysis is facilitated, and a richer view angle is provided for data analysis; in combination with a graph-based sorting algorithm for texts and an abstract text abstract model, encoding answer documents contained in answer document clusters of different answer document topics into answer document abstracts, and forming topic-abstract pairs with hierarchical relationships by the answer document topics and the answer document abstracts of the answer document clusters; manual intervention is reduced, human errors are avoided, the consistency and reliability of medical data processing are improved, and the medical data processing time is shortened.

The invention discloses a file disassembling method based on multi-dimensional file association analysis and automatic interpretation, and relates to the crossing field of natural language processing and big data analysis. Comprising the steps of (1) carrying out multi-source heterogeneous data collection, (2) carrying out intelligent cleaning on the collected data and constructing a knowledge graph, (3) carrying out deep correlation analysis, (4) carrying out automatic disassembly and file simulation, and (5) carrying out intelligent application and feedback: automatically generating a Markdown document containing a file abstract, clause comparison and an enterprise portrait through a structured report engine, and carrying out intelligent application and feedback on the Markdown document. GIS map overlay analysis is supported, clauses in an applicable file are matched based on an enterprise tag system, declaration materials are pre-checked through an OCR + rule engine, user feedback cases are collected, model parameters are incrementally trained monthly, it is ensured that adaptation is completed within 24 hours after a new file is released, and a collection-analysis-application-optimization closed-loop iteration mechanism is formed.

The invention provides a purchase review document abstract generation method and system combined with an AI large model, and relates to the technical field of large language models.The method comprises the steps that firstly, a set containing a plurality of purchase review document units with review time stamps is obtained, and the content theme feature and the review logic feature of each purchase review document unit are extracted; calling a pre-trained AI large model to carry out combined abstract generation processing on the content theme features and the review logic features to generate a preliminary abstract text, then carrying out semantic coherence optimization processing on the preliminary abstract text to obtain a semantic coherence optimized abstract text, and finally, carrying out semantic coherence optimization processing on the semantic coherence optimized abstract text. And generating a final procurement document abstract containing the theme keyword set based on the optimized abstract text, and outputting the final procurement document abstract to the target storage terminal according to a preset format, so that the procurement document abstract is automatically, efficiently and accurately generated, and the document processing efficiency and the information utilization value are improved.

The invention relates to an SSM-based model for generating long document abstracts / keywords. The SSM-based model comprises an encoder, a decoder and an output vector module, the encoder comprises an SSM and a bit-by-bit feed-forward network. The SSM is used for converting an input text sequence through hidden state representation to generate an output text association sequence; the bit-by-bit feed-forward network is used for performing nonlinear transformation on the representation vector of each position in the text association sequence to generate code output; the decoder is of a multi-layer structure, and a target sequence is generated step by step according to codes output by the encoder and output of a previous layer of the decoder; and an output vector module reads a target sequence output by the decoder, and obtains a text output by the model based on the temperature coefficient and a Top-k sampling method. By introducing the SSM, the model can process longer texts, and meanwhile, the expenditure of computing resources is reduced. And then, in a decoder, a decoding strategy of Top-k and a temperature coefficient is introduced, so that the content of the generated abstract text is more accurate and closer to the theme.

The invention discloses a method for automatically filing and managing group messages, and relates to the technical field of data management. Comprising the steps of (1) establishing a community enterprise WeChat group, (2) establishing a unified rule, and naming and publishing the same summarized or collected question in the WeChat group by adopting the same identifier according to the unified rule, (3) automatically obtaining various types of information in the WeChat group and synchronizing the information to a database, 4, analyzing and classifying the information obtained from the WeChat group by using an AI large model, generating a corresponding file abstract, and forming a file label according to the classification; 5, retrieving the information content in the WeChat group: retrieving by using a multi-dimensional list condition and / or a historical dialogue review mode according to a retrieval condition, the retrieval conditions comprise a time period, a file type, a file name, a sender and file content.

The invention provides a method for enhancing large model RAG retrieval efficiency. The method specifically comprises the steps of data index optimization, retrieval query optimization, retrieval result optimization and generation process optimization. During data index optimization, a large language model is adopted to process and mark data before data storage, data density is improved, low information amount and noise are reduced, and the method is used for reducing corresponding cost and improving accuracy. During data processing, a document abstract is created, and first-layer filtering is performed by using the document abstract, so that the retrieval efficiency is improved; a series of hypothetical / possible question and answer pairs are provided for each document, and then the generated questions are used as embedded retrieval blocks to improve the symmetry of retrieval; and then deleting repeated information by using a large language model, and simplifying data indexes and improving the response quality of the RAG system by clustering and merging the repeated information. According to the method, the retrieval efficiency and the retrieval accuracy can be further improved through various optimization schemes in actual use.

The embodiment of the invention provides a problem processing method and device based on a large language model, and the method comprises the steps: carrying out the text screening in a preset text set according to a user problem, obtaining an intermediate text, and carrying out the first filtering processing of the intermediate text based on the relevancy between the user problem and the intermediate text, and obtaining a first remaining text; inputting the first residual text and the user question into a large language model to generate an answer to obtain a question answer of the user question; performing second filtering processing on the intermediate text based on the relevancy to obtain second residual texts, and selecting at least one of the second residual texts and a preset document abstract according to the number of the second residual texts; and inputting at least one, the user question and the question answer into a large language model for question recommendation processing to obtain a recommended question. By adopting the embodiment of the invention, the accuracy of answers to the questions of the user can be improved, and the initiative of question and answer interaction of the user is improved.

The embodiment of the invention relates to a dense and sparse retrieval fusion method and device based on summary words, and the method comprises the steps: selecting a large language model as a working model, taking a prediction header / instruction template used by the model in a next word prediction task as a summary word prediction header / instruction template, and taking a model word list as a working word list; creating a document summary word dense vector mapping space, namely a first space, for the library by using the instruction template, the working model and the prediction header; receiving a user query; using the instruction template, the working model and the prediction header to identify the dense vector and the probability distribution vector of the queried summary word, and performing keyword identification according to the query, the probability distribution vector and the working word list; performing dense retrieval on the library according to the queried dense vector and the first space; performing sparse retrieval on the library according to the keyword sequence; and feeding back a fusion result of dense and sparse retrieval to a user. According to the invention, the calculation overhead can be reduced, and the retrieval precision is improved.

The invention discloses a knowledge graph multi-modal document analysis and image table semantization knowledge recall method, and belongs to the technical field of knowledge engineering and information retrieval. The invention provides an innovative scheme for fusing a visual language model, semantic abstract generation and knowledge graph modeling. The method comprises the following steps: constructing a vertical domain knowledge graph by adopting a BERT-BiLSTM-CRF model; according to the method, multi-modal document analysis is realized through models such as DocLayout-YOLO, TableMaster, UniMERNet and the like; the method comprises the following steps of: segmenting an image into 16 * 16 block sequences by adopting a vit-gpt2-image-adaptation model, and realizing image semantization through 768-dimensional vector space mapping and Transform coding; constructing a document summary tree based on DBSCAN clustering and LLM recursive summary; and designing a hybrid retrieval space fusing semantic vectors and structured vectors, and reordering by adopting a double-attention mechanism. According to the method, the knowledge base document retrieval recall rate is increased to 99%, the question and answer accuracy rate reaches 90% or above, the index construction time is shortened by 60%, and the problem that semantic understanding and recall of non-text elements in complex documents are difficult is effectively solved.

The invention discloses a method for constructing an intelligent search engine based on a large language model, comprising: step 1: original query; step 2: manual setting of filters; step 3: based on the original query and filter setting input by the user, the retrieval rewriting module is responsible for converting it into a professional retrieval statement, and the modules involved in the retrieval rewriting process include: dynamic few-shot examples based on NER, professional terminology recommendation, retrieval main sentence rewriting based on LLM, retrieval main sentence verification, filter automatic planning module, retrieval sentence conversion module and retrieval main sentence merge filter clause; step 4: using a retrieval engine to retrieve relevant documents; step 5: document reordering; step 6: document summary and indexing. The invention optimizes the query rewriting module and the reordering module, and the accuracy of retrieval rewriting and the rationality of recall document reordering are significantly improved.

The present application discloses a method for generating a document summary, including: encoding user intent information to obtain encoding information of the user intent information; encoding a target document to obtain encoding information of the target document; determining the importance of words contained in the target document corresponding to the target document and the user intent information according to the encoding information of the target document and the encoding information of the user intent information; adjusting the encoding information of the target document according to the importance; generating a document summary of the target document according to the adjusted encoding information. The above method is used to solve the problem that the document summary generated under the prior art cannot meet the personalized needs of users and reduces the user's product experience.

The invention discloses an interactive application security test vulnerability study and judgment method and system based on AI, and relates to the field of electric digital data processing, and the method comprises the steps: carrying out the instrumentation of a key function of a target system, and collecting execution environment information; and identifying user input from HTTP request data in the execution environment information, and tracking a propagation path of the user input to obtain a polluted data propagation chain with a vulnerability type mark. And packaging the execution environment information, the pollution data propagation chain, the vulnerability type mark and the related function call stack into a to-be-analyzed vulnerability information packet, and sending a system description document and the vulnerability information packet combination to a large language model to obtain a document summary. And the document summary and the pollution data initial entry point information are combined and sent to a model, and a context supplement request is acquired. And sending the supplement request and a preset vulnerability bypassing case combination to a model to obtain a research and judgment result. And finally updating the research and judgment result to the vulnerability detail information. By implementing the method, the verification cost of security test vulnerability research and judgment can be reduced.

The invention provides a document analysis method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a to-be-processed document, summarizing the to-be-processed document, and generating and storing a document abstract; after obtaining a target question of the to-be-processed document, performing intention recognition on the target question, and determining a request type corresponding to the target question; the request type comprises document summarization or document questions and answers; under the condition that the request type is a document summary, obtaining a target answer corresponding to the target question based on the document abstract; and when the request type is the document question and answer, obtaining a target answer corresponding to the target question based on the to-be-processed document. Therefore, corresponding target answers can be efficiently and accurately generated according to different request types, so that the effect of providing accurate answers according to specific requirements of the user is achieved.

The invention relates to the technical field of data processing, and discloses a text governance method and device, equipment and a medium, and the method comprises the steps: collecting a preset enterprise external rule file in real time through a preset double mechanism of a timing scheduling task and an event triggering task; screening out files which do not accord with a preset category, labeling the files according to level labels, and storing the files into an internal database to form an internal specification file; extracting an internal specification data abstract, calculating a correlation coefficient between each service line label and the abstract, and determining a service line corresponding to the internal specification; constructing a knowledge graph by taking the file, the abstract, the label and the level as nodes and taking the correlation coefficient as an edge; when external specification changes are monitored, change content is collected, corresponding internal specification target files are identified, service labels with correlation coefficients exceeding a threshold value are screened to form a set, and corresponding service line internal specifications are collected; and by taking the change external specification level as a level, updating the internal specification corresponding to the service label with the level lower than the level in the set according to the change content. The accuracy of text governance can be improved.

The present invention relates to the technical field of intelligent document extraction, and discloses an intelligent abstract extraction method and system for tender documents based on image recognition, including selecting the application field to which the tender document belongs according to the tender document, inputting the tender document information through image scanning, establishing an abstract extraction model to extract the keywords of the tender document, establishing a tender document abstract by sorting the keywords, and having a tuning and correction unit check whether the abstract is smooth and whether it conforms to the semantics of the original tender document. If the abstract is smooth, continue to check whether it conforms to the semantics of the original tender document. If the abstract is not smooth, perform synonym replacement and root word association on the keywords and re-sort the keywords. If it conforms to the semantics of the original tender document, complete the intelligent extraction of the tender document abstract. If it does not conform to the semantics of the original tender document, re-check whether the keywords are correct, realizing intelligent abstract extraction, providing a complete algorithm logic, and ensuring the accuracy of the extracted abstract of the tender document.

The present disclosure pertains to dynamic query classification and routing in multi-model AI (Artificial Intelligence) architectures. A method includes receiving a user query at a computing device, generating an embedding representation of the query using a pre-trained language model, and applying a trained classifier to the embedding representation to determine the query's type. The classifier categorizes the query into one of three types: information retrieval, document retrieval, or document summary. A response is then generated based on the classified query type, enhancing the system's ability to provide relevant and accurate results to users.

Described herein are examples of a system comprising a processor and memory storing instructions to execute a predictive multi-modal retrieval subsystem configured to analyze digital content, extract metadata, predict information needs, and retrieve relevant content; a multi-document viewing subsystem configured to index documents, establish relationships, and present aggregated content in a unified interface; a data management subsystem configured to organize files, generate folder structures, and provide interactive navigation; a file summary LLM configured to process documents and generate summaries with metadata; and a user context blob configured to maintain user context data across sessions. The system includes a method for receiving digital content, processing through OCR to extract text, analyzing with the file summary LLM to generate metadata and summaries, predicting information needs by simulating task progression and identifying related documents, and presenting retrieved content through a unified interface.

The present invention discloses a large file uploading method for a narrowband network environment, which belongs to the technical field of file uploading, and includes S1, loading a large file to be uploaded into a browser supporting HTML5 language to obtain first file summary information; S2, uploading the first file summary information, if the first file summary information already exists in the background database, notifying the browser to complete the upload immediately, if not, executing step S3; S3, calculating the slice size of the large file to be uploaded, and cutting the large file to be uploaded into corresponding block files in real time; S4, uploading the block files to the background server one by one, merging the block files after all the block files are transmitted, and completing the large file upload. Through the collaboration of the browser and the background program, the method realizes the reduction of large files to small files, and ensures the browser performance; it can realize breakpoint resume and second transmission, greatly improve the upload efficiency, make the operation simple and efficient, and also greatly improve the robustness of the system and the narrowband link adaptation ability.

The invention discloses a full-automatic matching method, device, equipment and medium for financial bills and receipts under a fuzzy condition, and the method comprises the steps: obtaining financial bill information and receipt information, and carrying out the feature extraction of the financial bill information and the receipt information, obtaining a bill characteristic value corresponding to each bill and a bill characteristic value corresponding to each bill; according to each document feature value, generating a document abstract corresponding to each document; according to the bill feature values, the bill feature values and the bill abstracts, matching the bills with the bills one by one, and outputting a first matching result; and if the unmatched bills and the unmatched bills exist, performing fuzzy matching according to the corresponding bill feature values and the corresponding bill feature values, and outputting a second matching result. According to the method, a two-step matching method is introduced, so that the accuracy of full-automatic matching of financial bills and receipts under fuzzy conditions is improved.

The present disclosure provides a document processing method, apparatus, electronic device, and storage medium, relating to the field of data processing technology. A specific implementation scheme comprises: obtaining a document collection, the document collection including multiple candidate documents, and determining segmented paragraphs and document summaries for the candidate documents; obtaining paragraph word vectors for the segmented paragraphs and summary word vectors for the document summaries, generating a vector index based on the paragraph word vectors and summary word vectors, and storing the index in a search engine cluster; and generating an inverted index based on the candidate documents and document summaries, and storing the index in the search engine cluster.

The invention provides a bidding file abstract generation method and system based on a large language model, and relates to the technical field of natural language process.The method comprises the steps that firstly, domain terms are extracted from an input to-be-processed bidding file set to construct a term library, and file units are split to form a domain semantic chain; inputting the term library into a pre-trained large language model to adjust the semantic analysis weight so as to obtain a semantic analysis model adaptive to the field; inputting semantic nodes in the semantic chain into the semantic analysis model to extract core expressions so as to form a dynamic abstract fragment set; determining an abstract fragment association sequence according to a semantic node connection relationship, and joining to obtain a preliminary fusion abstract text; and finally, semantic feedback of the reader is received, abstract fragment expressions and sequences are adjusted, a semantic chain is updated, and a final abstract is generated and output, so that the high-quality bidding document abstract can be efficiently generated.

The present invention provides a method for generating judicial document summaries based on three-stage GRPO reinforcement learning, which belongs to the field of data processing technology. The method specifically includes: step 1, modeling a three-stage thinking chain; step 2, performing data distillation and stratification on the original judicial document dataset according to the three-stage thinking chain to obtain different types of datasets, wherein the types include high correlation, medium correlation, and low correlation; step 3, using the high-correlation dataset to perform SFT supervised fine-tuning training on the large language model; step 4, using the entire dataset to perform multi-stage GRPO reinforcement learning training on the trained large language model to obtain a target model; step 5, inputting the target judicial document into the target model to generate a target summary. The solution of the present invention improves the efficiency, accuracy, and adaptability of summary generation.

The summary of a document describing an incident related to a specific information security vulnerability should include information about that incident. [Solution] The extraction unit extracts damage information, including keywords related to the content of the damage case, from a document describing an example of damage caused by a specific vulnerability in information security. The generation unit generates an explanation of the damage case using the damage information extracted from the document. The output unit outputs information based on the generated explanation.

The application discloses a kind of intelligent interaction method and system of document.The method includes receiving the document uploaded by user, generates document list;Receive the reading request of target document in the document list by user, generate conversational reading interaction interface, the conversational reading interaction interface includes tool area, document outline area, document current reading area and conversation window area;The document summary of the target document is generated and shown in the conversation window area by document assistant, and the document assistant includes ChatGPT natural language processing model;The content selection of the target document in the document current reading area by user is monitored in real time and the question of user is obtained;The answer of the question is generated by the document assistant and is shown in the conversation window area.The application provides a kind of document reading scheme that can be intelligently interacted, improves user reading experience, and can better meet the reading demand of user.

The present disclosure is directed toward systems, methods, and non-transitory computer readable media that generate a preliminary predicted document-summary consistency portraying elements from a text prompt utilizing a generation diffusion model and refine the preliminary predicted document-summary consistency to generate a predicted document-summary consistency. In particular, the disclosed systems receive, via an interaction with a user device, a text prompt specifying elements to portray within a predicted document-summary consistency. Furthermore, the disclosed systems generate an image generation prompt from the text prompt. Moreover, the disclosed systems utilize the generation diffusion model to generate a preliminary predicted document-summary consistency depicting the elements from the text prompt. In addition, the disclosed systems refine the preliminary predicted document-summary consistency to generate the predicted document-summary consistency.

The invention belongs to the technical field of natural language processing, particularly relates to a multi-source document clustering method, system and equipment based on machine learning, and aims to solve the problems of weak representation capability, poor semantic consistency, insufficient label interpretation and the like in document clustering. The method comprises the steps of obtaining a plurality of documents to be clustered; taking the document and the target document abstract word number as input data of a first language model, and generating a document abstract according with the target document abstract word number through the first language model; performing semantic analysis on the document abstract, and converting the document abstract into an embedded vector based on a semantic analysis result; dividing the document into a plurality of document clusters based on the similarity between the embedding vector of any document and the embedding vectors of other documents; generating a clustering label for representing a document clustering semantic topic; and establishing a document set under the same document cluster according to the cluster labels, and displaying the document set in a structured view. According to the scheme, the cross-domain document set can be effectively normalized and clustered.