81 results about "Word count" patented technology

The invention relates to the field of text processing, in particular to a retrieval system for a bidding law document large language model. Comprising a document preprocessing module, a semantic dicing module, an entity and relation recognition module, a knowledge graph construction module, a multi-strategy mixed retrieval module and a retrieval result preferential module. During working, the bidding law document is preprocessed, based on document word number judgment, a chapter-level slicing strategy or a clause-level slicing strategy is adopted for cutting, a LateChunking algorithm is used for determining cutting points, semantic units are extracted, entity elements are recognized, a knowledge graph is constructed, and an optimal retrieval result is obtained through multi-strategy mixed retrieval and preferential processing. According to the method, the limitation of a traditional single retrieval mode in bidding legal chief document processing is overcome, and the provision retrieval accuracy and context coherence are improved, so that the illusion risk caused by information missing or misunderstanding of a large language model is greatly reduced, and the credibility and practicability of a legal intelligent question and answer result are enhanced.

The invention provides an RAG method based on semantic precise blocking and precise background information generation, and belongs to the technical field of artificial intelligence. The method comprises the following steps: in a preparation stage before reasoning, preprocessing a large number of related knowledge texts, and segmenting the texts into coarse-grained text blocks according to the number of words; dividing the coarse-grained text block into a plurality of fine-grained text blocks according to semantics through a large language model, and generating a context abstract and an association problem for each fine-grained text block; splicing the fine-grained text block and the context abstract into a final text block, and taking the final text block as a block containing accurate background information; and storing the final text blocks and association problems in different vector databases after vectorization. In the reasoning stage, when facing a user question, the system firstly vectorizes a user question text, then retrieves the most relevant items from the different vector databases for merging, and then optimizes and sorts the retrieved results through a reordering model. And selecting the top-ranked text blocks for the large language model to generate answers. According to the method, the problems of inaccurate blocking, context information missing and inaccurate retrieval in the traditional RAG method are effectively improved.

The invention discloses a two-stage long text report generation method and system, and relates to the technical field of artificial intelligence. The invention provides a two-stage long text report generation system. The two-stage long text report generation system comprises an article outline generation module, a report generation module and an abstract and summary report generation module, the long text report generation method based on the system comprises the following steps: according to a report generation demand, performing theme investigation and view discovery, simulating a domain expert dialogue through an intelligent agent to guide a report theme, establishing a rich knowledge base, and generating a detailed article outline based on the knowledge base; according to the word number and style requirements of the user, the knowledge base and the article outline are combined, long text writing is carried out, reference literature numbers are embedded, generated report content and related literatures are exported as PDF, and the user can store and share the report content and the related literatures conveniently. Therefore, by the adoption of the method, the outline with strict logic can be generated, diversified personalized requirements are met, and the method is suitable for report requirements in different fields.

The invention discloses a local knowledge base RAG method based on Bayesian reasoning, and belongs to the technical field of information retrieval. The method comprises the steps of obtaining all knowledge data of a local knowledge base, performing paragraph classification and coding on the knowledge data to obtain a paragraph set, and calculating the semantic probability of each paragraph in the paragraph set; obtaining a target professional vocabulary set of the target question, and calculating the semantic probability of each vocabulary in the target professional vocabulary set in each paragraph and the semantic probability of all vocabulary in the target professional vocabulary set in each paragraph in the paragraph set Calculating the conditional probability of each vocabulary in the target professional vocabulary set in each paragraph based on the occurrence frequency; and calculating the conditional probability of the target professional vocabulary set in each paragraph by adopting a Bayesian formula, sorting to obtain a conditional probability sorting result, selecting the paragraphs of which the total word number does not exceed a preset threshold value as a target paragraph set, and generating target retrieval information based on the target paragraph set and the target question. According to the method, the accuracy and reliability of question retrieval are improved.

The invention discloses a multi-channel integrated marketing contact content management system and method, and relates to the technical field of marketing contact content auditing, and the method comprises the steps: creating a plurality of marketing contacts in a unified management platform, and configuring a delivery channel; establishing a design specification rule base; receiving a marketing content picture uploaded by the user and associating the target channel with the contact; calling a rule base on the basis of a contact type, starting an AI auditing engine to carry out automatic compliance inspection on the picture, including preprocessing, layout identification, character extraction and character number statistics, font and character size identification, color identification and functional element detection, and comparing an identification result with a rule to generate an auditing result; and approving the delivery of compliant content and tracking the effect, and optimizing an AI engine and a rule base based on data. According to the method, AI intelligent auditing is embedded into the marketing content publishing process, multi-channel integrated management and intelligent auditing of the marketing content are achieved, the auditing efficiency and accuracy are remarkably improved, it is guaranteed that the content is standard and unified, and the labor cost is reduced.

The invention relates to the technical field of information, in particular to an intelligent short message platform management and control method and system based on multi-level risk control. The method comprises the following steps: reading a to-be-sent short message text, comparing the to-be-sent short message text with a preset interception keyword library, and intercepting a to-be-sent short message if a matched interception keyword exists; identifying the semantics of the short message to be sent, and comparing the semantics with a plurality of preset prohibited scenes to obtain a scene matching degree; if the scene matching degree is higher than a preset threshold value, intercepting the short message to be sent; according to the recognized semantics, marking a text corresponding to the semantics; reading the unmarked text, and if the word number of the unmarked text exceeds a preset word number threshold value, obtaining characters with similar fonts and characters with similar pronunciation; trying to perform semantic recognition according to the combination of the characters with similar fonts and the characters with similar pronunciation of the unmarked text; comparing with a plurality of preset forbidden scenes; and if the matching degree of any second scene is higher than a preset threshold value, intercepting the short message to be sent.

The invention discloses a super-long text generation method and device, equipment and a storage medium, and the method comprises the steps: responding to a document generation instruction, and generating an initial text outline of a target text through a large model; wherein the initial document outline comprises a plurality of chapter titles; based on the keyword information of the initial document outline, the text complexity of the target text is obtained through calculation, and the text complexity is associated with the subject breadth of keywords and the expected word number of the target text; if the text complexity of the target text is greater than a preset threshold value, determining the initial text outline as a target text outline; and on the basis of the target text outline, the text content corresponding to each chapter title is generated in parallel and spliced to obtain the target text, so that the generation efficiency, the text coherence and the overall text quality of the generated super-long text can be improved, and the generation cost of the super-long text is reduced.

The invention discloses a volume and noodle retrieval matching method and device and a storage medium. The method comprises the steps that a volume and noodle library is constructed; acquiring a first volume image; identifying and erasing handwritten handwriting; and paper surface information matching. According to the method, the volume information of the template is retrieved in a text matching mode, and compared with simple image similarity matching retrieval, the speed is higher, and the accuracy is higher. Moreover, according to the method, the front surface and the back surface of the test paper are respectively retrieved, and the confidence coefficient of the retrieval result is calculated according to the text word numbers of the front surface and the back surface of the retrieval result, so that the accuracy and the credibility of the retrieval result can be improved, and the method can be suitable for retrieval and matching of the test paper surface information of more subjects.

The invention relates to a text slicing and recall method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a to-be-processed document which comprises at least one original paragraph; for each original paragraph, semantic segmentation is carried out on the original paragraph to obtain at least one text block, and the word number of the text block is smaller than or equal to the maximum word number limit; according to a title structure of the to-be-processed document, a title path of each original paragraph is extracted, feature extraction is carried out on the title paths, title feature vectors are obtained, and the title paths are used for indicating titles of each level to which the original paragraphs belong; and for each text block, taking the title feature vector of the original paragraph to which the text block belongs as an index field of the text block to obtain a text slicing result, and using the index field to retrieve the text block. In this way, the semantically related text blocks are associated during retrieval, and the situation that related information is dispersed is reduced, so that the retrieval accuracy is improved, and the reliability of generating answers by the RAG system is improved.

The invention discloses a self-adaptive large model paper review method and system, electronic equipment and a storage medium. The method comprises the following steps: acquiring an academic paper and review data thereof; the academic papers are input into a large language model for training, features are extracted from the text papers, and the features comprise the number of words, term density, syntactic complexity and semantic complexity; obtaining academic paper granularity information based on multiple levels of evaluation of word number, term density, syntactic complexity and semantic complexity; and obtaining review content based on the granularity information and related information in the knowledge base. The method can significantly improve the efficiency and quality of academic review, provides powerful technical support for academic research and teaching, and is expected to be widely applied in the academic and education fields.

This application proposes a language model training method, a text generation method, and related apparatus. The method allows the language model to predict the weights of each word in the original sentence, obtaining weight values ​​for each word. Each word's weight represents the semantic importance expressed by that word in the original sentence. Then, a training sentence is input into the language model, enabling it to generate text based on the training sentence. The language model's loss is determined based on the weight values ​​of the masked portion of the original sentence and the generated text, and the model's parameters are adjusted accordingly. This setup allows the model to distinguish between different words and adaptively adjust the semantic importance of each word, rather than indiscriminately counting incorrect words. This enables the language model to truly understand semantics and improve the accuracy of prediction results.

The present disclosure relates to a text processing method and device, equipment and storage medium, and belongs to the technical field of computer, which can improve the accuracy of text processing. The text processing method comprises: obtaining a to-be-modified text; determining a target text according to the to-be-modified text, a preset number of words and a pre-trained text modification model; the number of words of the target text is the preset number of words; the text modification model is obtained according to a sample text pair and a sample number; the sample text pair comprises a to-be-modified sample text and a modified sample text; and the sample number is used to represent the remaining number of words after the byte corresponding to the sample number in the modified sample text.

The invention relates to a text display and voice broadcast synchronization method and device, computer equipment, a readable storage medium and a program product. The method comprises the following steps: acquiring an original character string needing to be displayed on a current page of a client and / or an original character string already displayed on the current page of the client; punctuation marks in the original character string are filtered, and a pure character string corresponding to the original character string can be obtained; obtaining the word number of a read character string in the pure character string; then, according to the word number of the pure character string and the word number of the read character string, a broadcast demarcation point used for segmenting the read character string and an unread character string in the pure character string can be determined; and based on the document object model, positioning to the broadcast demarcation point in real time, and displaying the original character string corresponding to the broadcast demarcation point on the current page so as to realize synchronous display of character display and voice broadcast of the original character string.

Provided in the embodiments of the present disclosure are a text generation method, a text generation model training method, an electronic device, a storage medium, and a program product. The text generation method comprises: using input data of a user to generate a text generation request, and sending same to a server, wherein the text generation request carries theme information of text to be generated and word count requirement information of the text to be generated; and receiving text generated by a text generation model of the server in response to the text generation request, wherein the text is text which is generated by means of the text generation model by using the theme information in the input data as a theme and in which the number of generated words satisfies the number of words indicated by the word count requirement information, and the text generation model is obtained by means of training on the basis of a general generative basic model. The present solution can accurately generate text that satisfies a word count requirement.

The invention provides a commodity recommendation method and system based on multi-dimensional entity analysis and mixed retrieval, and belongs to the field of natural language processing. The method comprises the following steps: preprocessing a user problem to obtain a preliminary user optimization problem and a general suggestion; key entities of the preliminary user optimization problem are extracted, and commodity recommendation related problems are screened based on entity matching; obtaining a preference entity based on the user behavior record; performing semantic extension on the key entity and the preference entity of the commodity recommendation related problem to obtain an extended entity; the extended entities are spliced and recombined, and historical recommendation results are matched based on mixed retrieval of semantic retrieval and full-text retrieval; and aggregating the historical recommendation results, generating commodity recommendation key point statements based on a specified format and word number, and outputting the commodity recommendation key point statements and the general suggestions. Therefore, the defects of the conventional technology in the aspects of problem screening, preference analysis, retrieval, result presentation and the like are overcome, and the commodity recommendation efficiency and quality are comprehensively improved in multiple links.

The present invention relates to the technical field of large models, and discloses a method and system for dynamic word count control of text output of a large language model. By collecting data, high-quality question-answer pair data and question data are automatically generated based on the large model; the large model is fine-tuned using the question-answer pair data set to obtain a professional model; a reward model is trained using the question data set to score the content generated by the large model according to the word count; the parameters of the word count control model are constructed and optimized based on the question-answer pair data set and the reward model based on a reinforcement learning algorithm; when a user inputs a question and a word count requirement, the word count requirement is assigned to each chapter, and the word count control model is iteratively called in units of chapters to generate each chapter, and finally the entire article is spliced ​​together. The present invention can accurately control the output word count of the large model, has the advantages of accurate and controllable generation, thereby meeting the needs of specific application scenarios, avoiding publishing problems or format errors caused by too many or too few words, and enhancing user experience.

The embodiment of the application provides an electronic signature method, an electronic signature device, an electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: obtaining a pre-stored name text of a target object; performing text recognition on the pre-stored name text to obtain a total number of words of the pre-stored name text; obtaining a word number threshold value that can be displayed on a first signature page; if the total number of words is greater than the word number threshold value, performing pagination on the first signature page to obtain a plurality of second signature pages; obtaining an initial electronic signature of the target object on the second signature page; and arranging the initial electronic signatures of all the second signature pages to obtain a target electronic signature. According to the embodiment of the application, the total number of words of the pre-stored name text is used to distinguish the signature scene, and the accuracy of electronic signature recognition can be improved.

The invention discloses a subtitle processing method and device based on an editing template and computing equipment. The method comprises the following steps: acquiring any original subtitle input for a target editing template, and determining a template word number threshold value corresponding to the original subtitle; decomposing the original subtitles into a plurality of minimum semantic units; according to the semantic association degree between the minimum semantic units, aggregating the plurality of minimum semantic units so as to split the original subtitle into at least one semantic cluster; and extracting a target semantic cluster from the semantic clusters corresponding to the original subtitles according to the template word number threshold, and generating target subtitles according to the target semantic cluster. According to the scheme, the generated target subtitles are composed of the integer semantic clusters, and the semantic clusters have complete semantics, so that the semantic integrity of the displayed target subtitles is guaranteed, the phenomenon of semantic interruption of the subtitles is avoided, the subtitle display effect is improved, and the user experience is improved.

An information processing device (100) infers an answer to a query. A generation unit (110) generates a word count model (41) and a third-order tensor count model (42) by counting words with parts of speech obtained from a pre-learning text (21). A correction unit (120) generates a word deletion count (31) and a third-order tensor deletion count (32) by counting words with parts of speech to be deleted that are obtained from a deletion text (22). The correction unit (120) subtracts the word deletion count (31) from the word count model (41), and subtracts the third-order tensor deletion count (32) from the third-order tensor count model (42).

The present invention discloses a text extraction method, apparatus, device, and medium based on hybrid compression, comprising: obtaining a first text set of text to be compressed; semantically expanding and replacing paragraphs and paragraph keywords in the first text set to obtain a plurality of second text sets; compressing and trimming the paragraph sets of the first text set and the paragraph sets of the second text set using a hybrid compression ratio to obtain a plurality of third text sets; screening the plurality of third text sets to obtain a plurality of fourth text sets; and screening compressed extracted text corresponding to the text to be compressed from the plurality of fourth text sets based on the text to be compressed, the paragraph sets of the fourth text set, and the word count of the paragraph sets. The present invention belongs to the field of text compression. The present invention can ensure text compression while achieving the closest semantic approximation.

The invention discloses a text vectorization method and a vector database. The text vectorization method comprises the following steps: acquiring a divider sequence and a target text; segmenting the target text into a plurality of output text blocks according to the segmentation symbol sequence; collecting each output text block output by the character divider to obtain an output text block sequence, wherein the number of overlapped words of contents of adjacent output text blocks in the output text block sequence is greater than an overlapped preset value; and vectorizing each output text block in the output text block sequence to obtain a vector text block sequence corresponding to the output text block sequence. According to the text vectorization method, the retrieval efficiency of the target text is improved, and the steps of incremental updating of the text are simplified.

The invention belongs to the technical field of artificial intelligence, and relates to a voice recognition method and device, computer equipment and a storage medium, and the method comprises the steps: receiving an answer voice which is fed back by a customer and corresponds to a proposed question text; converting the answer voice into an answer text, and counting the word number of the answer text; judging whether the word number is less than or equal to a word number threshold; if yes, performing semantic analysis on the answer text and the question text based on the target large language model to obtain a correlation score between the answer text and the question text; judging whether the correlation score is greater than or equal to a preset score threshold value; if yes, performing knowledge verification on the answer text based on the knowledge base to obtain a knowledge verification result; and generating a voice recognition result corresponding to the answer voice based on the knowledge verification result. In addition, the voice recognition result can be stored in the block chain. The voice recognition method and device can be applied to insurance case reporting scenes in the financial field and the medical health insurance field, and the accuracy of voice recognition of the answer voice is effectively improved.

Disclosed are various embodiments for methods and systems of live caption display. Various embodiments can obtain a captioning file comprising a first caption and a second caption. Then, various embodiments can obtain a live transcript from a transcription service, the live transcript representing text generated from audio input. Various embodiments can then send a first instruction to a client device to display the first caption. Then various embodiments can calculate a first match percentage, the first match percentage representing a first amount that the first caption matches a first portion of the live transcription over a first word count of the first caption. Then, various embodiments can send a second instruction to the client device to display the second caption.

A system includes at least one processor to ingest a corpus of a plurality of documents that comprises training data, parse each document in the corpus of the plurality of documents to determine a word count and a raw risk for each document, determine a normalized risk for each document in the corpus of the plurality of documents using the word count and the raw risk based on an expected mean and an expected standard deviation based on a power risk that equals raw risk0.39 for the corpus of the plurality of documents, receive a new document, parse the new document to determine a word count and a raw risk, and determine a normalized risk for the new document based on the expected mean and the expected standard deviation based on the power risk that equals raw risk0.39 for the corpus of the plurality of documents.

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.

Methods and systems provide for presenting the results of talking speed analysis per topic segment in a communication session. The system connects to a communication session with a number of participants; receives a transcript of a conversation between the participants produced during the communication session, the transcript including timestamps for each utterance of a speaking participant; determines, based on analysis of the transcript, a meeting type for the communication session; generates, based on the meeting type, a number of topic segments for the conversation and respective timestamps for the topic segments; for each topic segment, determines a word count per unit of time for each speaking participant associated with a prespecified organization; and presents, to one or more client devices, data on the topic segments, each data including the word count per unit of time for each speaking participant within the topic segment.

The invention relates to the technical field of speech synthesis, in particular to an SSML visual editing method and security verification system based on packaging element mapping, and the method comprises the steps: S1, packaging an SSML native label into a customized DOM packaging element; s2, capturing a cursor position during user interaction, and dynamically inserting the DOM packaging element into the cursor position; s3, verifying the user input internal capacity through a real-time security verification mechanism, and constraining the user input internal capacity as an effective value domain; s4, synchronizing a user operation result to the attribute of the original label of the bottom-layer SSML, and generating a standard SSML document; and S5, dynamically stripping all DOM packaging elements and internal SSML labels in the text stream, and counting the number of characters of the remaining plain text content. According to the method, the SSML labels are packaged into the visual packaging elements, visual editing, accurate character level control, real-time safety verification and accurate word counting are achieved, and the efficiency, safety and user experience of SSML editing are remarkably improved.

An example method of language detection includes: for each word of at least a subset of words of a document, identifying a primary natural language associated with the word; associating each natural language of the set of natural languages with a corresponding word count indicating a number of words of the subset of words for which the natural language has been identified as the primary natural language; identifying, among the set of natural languages, a natural language associated with a maximum word count; and associating the identified natural language with the document.

The invention discloses a file data transmission method, device and system based on an ARINC 429P3 specification, and relates to the technical field of avionics system communication. The method is applied to a destination end, and comprises the following steps: receiving a request to send an RTS protocol word, the RTS protocol word comprising a first word count field; determining whether to accept the transmission request based on the word count of the first word count field and the residual capacity of the receiving buffer area of the destination end; in a case where it is determined to accept the transmission request, the transmission may send a CTS protocol word, where the CTS protocol word includes a second word count field having a word count less than or equal to a word count of the first word count field. According to the invention, the transmission efficiency when the receiving cache of the destination end is temporarily full can be effectively improved.

The embodiment of the invention provides a topic mining model training method and device and a storage medium, and can be applied to the technical field of natural language process.The method comprises the steps that for any word segmentation category in multiple obtained word segmentation categories, whether the word segmentation category meets semantic consistency or not is determined through a large language model; determining corresponding topics and word counting matrixes of the plurality of sample documents under the word segmentation categories with the semantic consistency, and determining global weights of the plurality of topics; each segmented word category corresponds to one theme, and the number of times of occurrence of the same segmented word in the plurality of sample documents is used as the word count of the segmented word under the corresponding theme; taking the topic and word counting matrix as initial values of topic and word distribution in a topic mining model; the topic mining model is trained through the plurality of sample documents and the global weight until the topic mining model with the target value and the word distribution is obtained, so that through semantic consistency judgment, the accuracy of topic mining is improved, and the training time of the model is shortened.