37 results about "Reference Document" patented technology

The invention discloses a document revision method, a document revision device and a storage medium, and relates to the technical field of data processing. The method comprises the following steps: firstly, acquiring a reference document and a to-be-revised document in a revision mode; and analyzing the reference document, and extracting structured revision information corresponding to the reference revision operation trace. And analyzing the to-be-revised document to generate structured document information containing target content and target format information. And then, generating a cue word based on the semantic intention of the structured revision information, and inputting the cue word and the structured document information into the target model to obtain target revision information. And finally, mapping the target revision information to the to-be-revised document, and generating a target document with a target revision operation trace. According to the document revision method, the normalization and accuracy of the revision result can be ensured, the manual intervention cost is reduced, and format errors and semantic errors are reduced.

The invention relates to the technical field of document management, and relates to a legal document key point automatic extraction and association method and system.The new legal document data of a target case is obtained, a preset document fingerprint algorithm is used for automatically comparing the new legal document data with a historical reference document, differential text data fragments are extracted, and the text data fragments are extracted; the method comprises the following steps of: performing deep semantic analysis on different text data fragments to generate candidate triple data containing entities and relationships, intelligently mapping the candidate triple data into a pre-constructed case knowledge graph, quickly positioning graph anchor nodes with data change, and then, performing data change on the basis of change types of the different text data fragments. And executing atomic updating operation on the map anchor node, including updating the dynamic mark of the node state attribute, and finally generating the change source node data with the latest state identifier. The problems that the current legal case data updating depends on manpower, is low in efficiency and is easy to make mistakes are effectively solved.

Embodiments of the present disclosure are directed to generating content filer forms from a set of documents uploaded to a cloud-based collaboration environment. Generating content filer forms from a set of documents can comprise uploading the set of documents, pre-processing the uploaded documents to determine one or more document types in the set of documents and an intent for each determined document type, generating a natural language prompt for each document type based on the intent for each document type and one or more reference documents defining constraints on the natural language prompt, generating, from each natural language prompt a content filer form associated with each document type using a generative Artificial Intelligence (AI), and refining each generated content filer form.

The invention provides an evaluation set generation method, equipment and a product. The evaluation set generation method comprises the steps that an RAG knowledge base containing knowledge documents and a question set of evaluation questions are acquired; extracting a theme of the knowledge document, and determining a document identifier of the knowledge document; constructing a first evaluation data set by using the theme and the document identifier; adding the question types into the first evaluation data set by utilizing a matching relationship between themes in the first evaluation data set and the question types in the question set to obtain a second evaluation data set; and inputting the second evaluation data set into the large model, and generating an evaluation set containing the document identifier, the theme, the question type, the question, the answer and the reference document.

The invention provides a model training method, a question answering method and a related device, and relates to the technical field of machine learning. The method comprises the steps of firstly obtaining a first sample data set and a reference document set; then a prompt text is constructed based on the first sample data set and the reference document set, the prompt text is used for prompting a machine learning model to select a reference document used for replying the first question data from the reference document set, and first answer data used for replying the first question data is generated; and finally, training a machine learning model according to the prompt text to obtain a question and answer model. Therefore, the machine learning model is trained in a thinking chain enhancement mode according to the prompt text, how to select related reference documents from the reference document set according to given questions can be learned, and key information is extracted to generate accurate answers; therefore, the training efficiency of the model is improved, the accuracy of the question and answer task is improved, and the trained model can adapt to large-scale data processing requirements.

This invention relates to the fields of artificial intelligence and software testing technology, and provides a content compliance review method for a "testing brain" based on KG and RAG collaborative knowledge enhancement. The method includes: acquiring a set of reference documents and obtaining the knowledge graph and high-dimensional vector corresponding to each reference document to construct a knowledge base; receiving a report to be reviewed; on the one hand, comparing the similarity between the retrieval vector of the report to be reviewed and the high-dimensional vector of the knowledge base to retrieve highly relevant reference documents from the knowledge base; on the other hand, selecting highly relevant subgraphs from the knowledge graph based on the entities in the report to be reviewed; fusing the highly relevant reference documents and highly relevant subgraphs to form structured knowledge, and generating structured prompts based on the structured knowledge; and outputting the compliance review results of the report to be reviewed based on the structured prompts and structured knowledge. This invention can significantly improve the efficiency, accuracy, and interpretability of software testing compliance review.

An information processing apparatus comprises: an acquisitor to obtain document information, including character strings and position data from document image data; a converter to transform the acquired document information into a distributed representation; an information extractor to identify a character string corresponding to an item specified by a prompt, using a large language model by inputting the prompt with the acquired document information; a storage unit to save the document information, distributed representation, and extraction results, associating them with the document; and a selector to choose a reference document from previously processed documents based on distributed representations. The prompt for processing a new document includes details about the selected reference document.

The invention provides a document analysis result evaluation method and device and electronic equipment, and the method comprises the steps: obtaining a reference document and an analysis document corresponding to a target document, each of the reference document and the analysis document comprising a plurality of document tags and at least one standardized tag under a standardized framework; according to the document label and the standardized label in the reference document, a first data report is generated, the document label and the standardized label in the document are analyzed, a second data report is generated, each of the first data report and the second data report comprises a plurality of report items, and each report item is used for describing document content from one dimension; and determining an evaluation result of the parsed document according to the first data report and the second data report. The analysis document can be evaluated from multiple dimensions, so that overall and integral evaluation results are obtained.

The invention discloses a question and answer retrieval method and device, and the method comprises the steps: obtaining a target question, and determining a plurality of sub-questions included in the target question; determining a first group of document slices matched with keywords of the target question and the plurality of sub-questions and a second group of document slices matched with semantics in a preset document slice library; and taking an intersection of the first group of document slices and the second group of document slices to obtain a reference document, and inputting the target question and the reference document into a preset language model to obtain a target answer for the target question.

The invention relates to a document processing method and device, equipment and a storage medium. The method comprises the steps of obtaining a reference document, and performing paragraph segmentation on the reference document according to a document structure of the reference document to obtain first metadata of each paragraph in the reference document; positioning a paragraph to be verified in the reference document according to the first metadata and the existence condition of a historical version document of the reference document; and according to the target paragraph content of the paragraph to be checked, searching a to-be-processed document from the historical document, and according to the target paragraph content, revising the to-be-processed document. By adopting the method, the document processing efficiency and the accuracy of a document result can be improved.

The invention discloses a long document conflict comparison method based on a multi-modal large model, and belongs to the technical field of document processing. The method aims at solving the technical problems that existing document conflict comparison can only recognize literal conflicts in a shallow layer, an NLP small model is poor in long text processing effect, and when a large model processes a long document, the computing power cost is high, semantics are prone to being lost, and compression is not thorough. According to the technical scheme, the method is characterized by comprising the steps of receiving, analyzing and converting a to-be-compared document and a reference document into a standardized format, splicing a text to a cue word and performing Token processing, executing three-level progressive dynamic compression if the text exceeds the limit, inputting the processed cue word into a multi-modal large model, completing literal and semantic layer conflict comparison by the model, and finally obtaining a comparison result. If the compression is invalid, executing a batch and document splitting bottom solving strategy, and finally formatting and outputting a comparison result.

A computer system may obtain a fact extraction prompt. The fact extraction prompt is configured to control how a fact generating machine learning model extracts or summarizes content of reference documents included in a corpus of documents. The computer system may input, into the fact generating machine learning model, the fact extraction prompt and one or more reference documents from the corpus of documents to identify one or more facts included in the one or more reference documents, populate respective data fields of one or more fact objects based upon the one or more facts identified by the fact generating machine learning model and generate one or more fact summaries for the matter based on the one or more fact objects. The one or more fact summaries include structured presentations of at least some of the one or more fact objects according to contents of the data fields.

The invention relates to a content generation method and device. The method comprises the steps of receiving a content generation instruction in a natural language form and a specified reference document; generating an outline structure according to the content generation instruction and the reference document; according to the content of each node in the outline structure, obtaining document fragments related to each node in a preset document library; and based on the outline structure, the document fragments related to each node and the content generation instruction, generating target content through an artificial intelligence model. Therefore, a two-stage intelligent creation technical scheme based on the document library is realized, the natural creation logic that a creator lists a creation outline first and then supplements chapter contents is simulated through ordered connection of an outline generation stage and a content generation stage, the problem that the creation contents are different from user themes and core requirements is effectively solved, and the creation efficiency is improved. The creation convenience and content quality are remarkably improved, and efficient, convenient and intelligent creation auxiliary services are provided for creators.

The invention discloses an information pushing method and device, a storage medium, a program product and computer equipment. The method comprises the following steps: acquiring a target document containing a user question and a reply content thereof; determining a first word set of the target document; based on a reference document and the difference between the time information of the reference document and the time information of the target document, determining first importance information of each word in the first word set; based on the target document, the first word set and the first importance information, determining a keyword set from the first word set; generating recommendation information based on the keyword set and a preset entry set; and pushing the recommended information, so that the user demand can be flexibly adapted, and the user experience of information pushing is further improved.

The invention relates to the technical field of artificial intelligence, and discloses an automatic document duplicate checking and rewriting method and device and a program product, and the method comprises the steps: obtaining a target document, carrying out duplicate checking comparison on the target document according to a reference document, generating a corresponding document similarity, and carrying out the matching of terms in the target document according to a user-defined term library. Proprietary terms in the target document are determined, other contents except the proprietary terms in the target document are rewritten according to the language model, and a rewritten document is obtained. According to the method, the corresponding document similarity is generated through automatic duplicate checking comparison, the special terms in the document are protected, then other contents except the special terms in the target document are rewritten according to the language model, and the two steps of duplicate checking and rewriting are connected, so that the number of times of repeatedly switching tools by a user is reduced, the labor cost and the knowledge cost are reduced, and the user experience is improved. The process of duplicate checking and rewriting is shortened, and the efficiency of duplicate checking and rewriting is improved.

The invention provides an information processing apparatus, an information processing method, a computer program product, and a recording medium. The information processing device includes: an acquisition unit that acquires document information included in image data of a document; a conversion means for converting the acquired document information into a dispersed representation; an information extraction unit that inputs a cue including the acquired document information into the large-scale language model, performs reasoning based on the large-scale language model, and extracts a character string corresponding to an item indicated by the cue; a storage unit that associates and stores in a storage unit document information relating to the document, the dispersion performance, and the extraction result by the information extraction unit; and a selection means for selecting a reference document from the documents that have been processed in the past on the basis of the distributed representation relating to the document that has been processed in the past and the distributed representation relating to the documents that have been processed in the past and stored in the storage unit, the presentation input to the document that has been processed including information relating to the selected reference document.

The invention provides a multi-document dynamic compression method and device, electronic equipment and a storage medium, and relates to the field of artificial intelligence such as large models, natural language processing and intelligent questions and answers. The method comprises the steps that an original query and K original documents are obtained, K is a positive integer larger than 1, and the original documents are reference documents, used for generating answers corresponding to the original query, of a target large model; obtaining importance evaluation results of the original documents respectively; according to an importance evaluation result, respectively determining a dynamic compression ratio of each original document; according to the dynamic compression ratio, content compression is conducted on the original documents, document fragments corresponding to the original documents are obtained, the document fragments are spliced, and a needed target compressed document is determined according to the splicing result. By applying the scheme provided by the invention, the quality of the obtained target compressed document can be improved, and the like.

The system uses a generation AI to modify document data and generate a database. The information processing device includes: an identification unit that identifies a search range for the database based on the attributes of the document data to be processed or the analysis results obtained by analyzing the content of the document data to be processed; a search unit that searches for document data corresponding to the document data to be processed from the identified search range; a modification unit that instructs the generation AI to modify the document data to be processed using prompts that designate the document data to be processed as the document data to be modified and the reference document data generated based on the searched document data as the document data to be referenced during modification; and a storage unit that updates the database by adding information based on the analysis results to the document data to be processed modified by the generation AI and storing it in an area within the database corresponding to the attributes of the document data to be processed.

The invention discloses a non-reference document image quality evaluation system. The system comprises a layout fusion downsampling module, a backbone network, a feature fusion structure and a quality regression module. And the layout fusion down-sampling module takes the document image and the layout prior information corresponding to the document image as input to obtain an optimized down-sampling result fusing the layout structure prior knowledge of the document image and the basic visual information. The backbone network takes an optimized down-sampling result as input, and semantic features, including bottom-layer features, middle-layer features and high-layer features, of a document image are extracted layer by layer through a multi-layer structure. And the feature fusion module fuses the image semantic features extracted from each layer of the backbone network and outputs multi-layer fusion features. And the quality regression module takes the multi-layer fusion features as input to generate multi-dimensional document image quality evaluation values. According to the method, the multi-dimensional quality of the document image can be accurately and efficiently evaluated, and the evaluation precision and the calculation efficiency in a complex distortion scene are improved.

A method of adjusting a document style includes obtaining a target document from a user, obtaining a style reference document; extracting, from the style reference document, document style information representing a document style of the style reference document, adjusting a document style of the target document, based on the document style information and a first external input signal indicating a first document style adjustment level of the target document, and displaying the adjusted target document with the adjusted document style.

The invention provides an online document processing method and system integrating a form control and change tracing in the technical field of intelligent cooperative processing of enterprise-level standardized documents. The method comprises the steps that S1, a reference document is obtained and analyzed to obtain a document structure, and a document template, a configuration directory function and an interactive form control are generated based on the document structure; s2, editing operation of adding or deleting the text content is carried out in the text area, a text identifier is generated for the text content in the editing operation process, a text comment is generated for the text content or the corresponding text content is copied, and filling operation is carried out on an interactive form control in the text area; s3, detecting the text content, and displaying a text identifier, a change record and a detection result in a functional area; and S4, correcting the text content based on a detection result in combination with a directory function, and exporting the corrected text content as a local document. The method has the advantages that the normalization, controllability and efficiency of document processing are greatly improved.

The invention relates to the technical field of text generation, and discloses a text generation method and system oriented to a strong fact constraint demand in the mine field, and the method comprises the steps: constructing a mine field knowledge graph, and defining core entities in the mine field and a core relation between the entities in the knowledge graph, constructing a core fact set based on the core relationship between the core entities in the mine field, and establishing an exclusive fact library based on the core fact set; obtaining a candidate retrieval document corresponding to user query, performing named entity recognition on the candidate retrieval document to obtain document key entity information, calculating a matching degree between the document key entity information and the exclusive fact library, and screening out a reference document based on the matching degree; integrating the reference document into a structured cue word, inputting the structured cue word into a model to output an initial text, and performing posterior verification and error correction on the initial text to obtain a final text conforming to strong fact constraints in the mine field; according to the method, the problem of fact deviation existing in an existing text generation mode is solved.

Provided is a method for predicting retrieval failure, the method being performed by a computing system. The method may comprise: selecting one of a plurality of documents in a document corpus as a reference document; selecting one or more positive samples having high relevance to the reference document from among the plurality of documents in the document corpus, using a retrieval model; selecting one or more negative samples having low relevance to the reference document from among the plurality of documents in the document corpus, using the retrieval model; calculating a gradient norm of a contrastive loss using the reference document, a positive sample, and a negative sample; and comparing the gradient norm with a first reference value, and predicting whether the reference document is a retrieval failure-causing document, based on a comparing result.

The invention relates to a document processing method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a document reference library and a document gray library about a target product; based on the document information of each gray level document in the document gray level library and the document information of each reference document in the document reference library, information comparison is carried out, target gray level documents belonging to a target change type in the document gray level library are determined, and the target change type is a modified type or a newly added type; determining change content in the target gray level document; and calling an editing specification tool to perform editing specification verification on the target grayscale document, and updating the document reference library based on the change content under the condition that the editing specification verification is passed. In this way, the document processing effect can be improved by saving the labor cost and improving the processing accuracy.

The invention discloses a two-stage document filtering and robust fine tuning method based on a graph attention network, and relates to the technical field of natural language processing and deep learning. The problems that an existing retrieval enhancement generation system is difficult to accurately identify useful documents in a mixed document environment and is easily interfered by anti-fact information are solved. The method comprises the following steps of: constructing a multi-class document pool comprising correct documents, anti-fact documents, noise documents and irrelevant documents, and constructing a semantic graph on a paragraph level; the method comprises the following steps: sequentially filtering irrelevant documents and noise documents by adopting a two-stage graph attention network to obtain a reference document set, constructing a document discrimination training sample and a question and answer training sample based on the reference document set, and combining the document discrimination training sample and the question and answer training sample into joint fine tuning data to perform instruction fine tuning on a large language model. And the model has a document reliability discrimination capability and keeps stable output in a mixed document situation, so that the factuality and credibility of a retrieval enhancement generation system in an anti-fact attack and noise environment are improved.

A two-stage document filtering and robust fine-tuning method based on graph attention networks is proposed, relating to the fields of natural language processing and deep learning. It addresses the problems of existing retrieval augmentation generation systems struggling to accurately identify useful documents in mixed document environments and being susceptible to counterfactual information interference. This method constructs a multi-class document pool including correct documents, counterfactual documents, noisy documents, and irrelevant documents, and builds a semantic graph at the paragraph level. A two-stage graph attention network is used to sequentially filter irrelevant and noisy documents to obtain a reference document set. Based on this reference document set, document discrimination training samples and question-answering training samples are constructed. These two sets are combined into joint fine-tuning data to fine-tune a large language model, enabling the model to possess document reliability discrimination capabilities and maintain robust output in mixed document scenarios. This improves the factual accuracy and credibility of the retrieval augmentation generation system under counterfactual attacks and noisy environments.

This specification provides a query task processing method and a document question-answering method. The query task processing method includes: obtaining document query information and multiple candidate document fragments corresponding to the document query information; constructing classification hints for the candidate document fragments based on the document query information and the candidate document fragments, and inputting the classification hints into a task processing model to obtain classification results for the candidate document fragments; using the classification results of multiple candidate document fragments, selecting a target document fragment from the multiple candidate document fragments; and generating a document query answer for the document query information based on the target document fragment. By constructing a binary closed candidate document fragment classification task, the task processing model can directly determine which candidate document fragments are answer reference documents related to the document query information, avoiding the establishment of a mapping relationship between the model response and the expected query result, thus improving the accuracy of query task processing.

To provide an information processor capable of accurately extracting a character string corresponding to a desired item from image data of a document.SOLUTION: The information processing apparatus includes acquisition means for acquiring document information (character string and position information) included in image data of a document, conversion means for converting the acquired document information into a distributed representation, extraction means for extracting a character string corresponding to an item instructed by the acquired document information by inputting a prompt including the acquired document information to a large-scale language model and performing inference by the large-scale language model, storage means for storing the document information related to the document and the extraction result by the information extraction means in a storage unit in association with each other, and selection means for selecting a reference document from previously processed documents based on the distributed representation related to the document to be processed and the distributed representation related to the previously processed document stored in the storage unit.SELECTED DRAWING: Figure 2

The invention discloses a file generation method and device, a file updating method and device, equipment, a medium and a program product in a cross-document reference scenario, and the file generation method comprises the following steps: obtaining other storage region parameters; the saved area parameter is used for indicating the position of a referenced target area in a referenced document; obtaining document content corresponding to the target area in the referenced document on the basis of the additionally stored area parameters; generating a target file based on the document content corresponding to the target area in the referenced document and the saved area parameters; the target file is used for updating the reference document based on the document content corresponding to the target area.