124 results about "Word count" patented technology

The invention provides a method and a system for generating an accurately referenced article based on a large model, and belongs to the technical field of article generation. The method comprises the following steps: acquiring an article generation demand input by a user, and generating an abstract of an article according to the demand by using a large model; retrieving knowledge related to the abstract in an RAG knowledge base based on the abstract; generating a series of question and answer pairs by utilizing a large model based on an RAG retrieval result; generating an outline of the article according to the title, the keyword and the question in the question-answer pair; performing importance evaluation on each chapter in the outline by using a scoring model, and calculating the number of words to be distributed to each chapter; splitting the generated outline according to the lowest-level title, combining the title, the keyword, the answer in the question and answer pair and the number of words allocated to the section, and inputting into a large model to generate corresponding text content; and combining the text contents according to an outline sequence to obtain a final article. According to the invention, word number control and accurate reference during article generation can be realized, the accuracy and reliability of article generation are improved, and the user experience is improved.

The invention discloses a text segmentation method and device for retrieval enhancement generation, and the method comprises the steps: constructing a text tree for an original text, and enabling the text tree to take an original text attribute as a root node, a title of each level as a child node, and a non-title paragraph as a leaf node; according to the method, text blocks are segmented according to a text tree, so that each segmented text block starts from a directory structure, ends with a non-title paragraph and is constrained by preset text block word number limitation, and the directory structure included in each text block is as follows: starting from a head leaf node included in each text block, starting from a non-title paragraph included in each text block, and ending with a non-title paragraph included in each text block; according to the method and the device, the text can be segmented in a mode more suitable for a tree structure, and text information loss caused in the text segmentation process is reduced, so that the recall effect of an RAG system is improved.

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 relates to the technical field of large models, and discloses a large language model text output dynamic word number control method and system, and the method comprises the steps: collecting data, and automatically generating high-quality question and answer pair data and question data based on a large model; performing fine tuning on the large model by utilizing the question-answer pair data set to obtain a professional model; training a reward model by using the question data set, and scoring the content generated by the large model according to the number of words; constructing and optimizing parameters of the word number control model based on a reinforcement learning algorithm by utilizing the question and answer pair data set and the reward model; when a user inputs questions and word number requirements, the word number requirements are distributed to each chapter, the word number control model is iteratively called with the chapter as the unit to generate each chapter, and finally the whole article is generated through splicing. According to the method, the output word number of the large model can be accurately controlled, and the method has the advantages of accurate and controllable generation and the like, so that the requirements of specific application scenes are met, publishing problems or format errors caused by too many or too few word numbers are avoided, and the user experience is enhanced.

The invention discloses a method and a device for automatically generating a PPT (Power Point) by an existing document based on a multi-modal large model, and belongs to the field of PPT automatic generation. The method comprises the following steps: S1, constructing PPT template pages, each template page comprises a plurality of shapes, and the fillable shapes are called elements; shape names corresponding to the elements are modified according to naming requirements, and PPT template construction is completed; s2, PPT text content is generated based on the document, document preprocessing is conducted firstly, and then the document is processed through a document self-adaptive processing mechanism; and S3, after the PPT template and the PPT text content are prepared, combining the PPT template and the PPT text content by utilizing an adaptive filling template design, filling the text content into a corresponding template position, and generating a key point content illustration. The method is not limited by the word number of the Word document and the number of the Word document, non-text information such as tables and pictures can be processed, manufacturing is easy, and the method is more attractive and reasonable.

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 relates to the technical field of data analysis, in particular to an intelligent interaction robot control system which comprises a voice receiving module, a text conversion processing module, an audio analysis processing module and an emotion judgment adjusting module. Wherein the text conversion processing module is used for converting purified audio signals into text content and determining text word length, adjective ratio and negative word ratio, the audio analysis processing module is used for determining the speech speed, fundamental tone characteristic trend and frequency fluctuation tendency of a single purified audio signal, and the emotion judgment adjusting module is used for determining character emotion tendency and audio emotion tendency. Determining the current emotional state of the user in combination with the character emotional tendency and the audio emotional tendency so as to determine whether to switch the voice output style of the interaction robot; according to the intelligent interaction robot control system provided by the invention, efficient voice processing and emotion analysis functions are integrated, so that the intelligent level of man-machine interaction is remarkably improved.

The embodiment of the invention provides a text generation method, a text model training method, electronic equipment, a storage medium and a program product. The text generation method comprises the steps that input data of a user is used for generating a text generation request, the text generation request is sent to a server side, and theme information of a to-be-generated text and word number requirement information of the to-be-generated text are carried in the text generation request; receiving a text generated by a text generation model of the server in response to the text generation request, the text being a text which is generated through the text generation model by taking theme information in the input data as a theme and whose word number satisfies the word number indicated by the word number requirement information, the text generation model is obtained based on training of a general generation type basic model. According to the scheme, the text meeting the word number requirement can be accurately generated.

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 provides a method and device for generating a research report based on a generative large model and a storage medium, and relates to the technical field of artificial intelligence and the like. According to the specific implementation scheme, the method comprises the steps of obtaining a title and description information of a research report based on obtained input information of a user; the description information is used for limiting the research report to be generated; obtaining knowledge information based on the title and the description information; based on the title and the description information, generating a target outline of the research report by adopting a pre-trained generative large model and adopting an at least two-step generation mode; the target outline comprises a plurality of chapter titles with a sequential relationship and target word numbers of chapter contents; and based on the target outline and the knowledge information, adopting the generative large model to generate the text content of the research report. According to the technology disclosed by the invention, the generation efficiency of the research report can be effectively improved.

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 relates to the field of data processing, in particular to a book recommendation method and system based on artificial intelligence, and the method comprises the steps: carrying out the classification and word segmentation of an article, obtaining words and the types of the words, and calculating the importance degrees of the words; generating a recommendation list according to the degree of importance of the words, and obtaining the reading time when the user browses articles in the recommendation list; dividing the reading time by the word number of the article as the reading speed of the user; calculating the interest degree of the user in the recommendation list; and according to the interest degree, obtaining a word related evaluation, and in response to the refreshing instruction, updating the recommendation list according to the word related evaluation. According to the method, the book recommendation accuracy can be improved, and the requirements of the user can be better met.

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.

The present invention proposes a method and system for text recognition and error correction, which performs pixel recognition on an image to be recognized and extracts the border of the text to be recognized based on a pixel processing algorithm to obtain a maximum text border; performs border matching on the maximum text border based on a preset library text frame to obtain text frame data; performs mapping processing on the text frame data based on the text data corresponding to the preset library text frame to obtain a number of text data; performs weighted scoring on the several text data based on the number of characters to be recognized and the text to be recognized to obtain a number of text weighted scores, and recognizes and corrects the text to be recognized. The present invention solves the problem that the recognition and error correction algorithms in the prior art are not effective, and usually misrecognition occurs due to the complexity of the algorithm, resulting in a low accuracy rate of optical character recognition. The present invention realizes pixel recognition and multi-level analysis of text, simplifies the recognition algorithm, and improves the accuracy rate of optical character recognition.

The embodiments of the present application provide a method for generating audio data, an instant communication method, and related devices. The method comprises: storing a received text stream into a text buffer word by word in the order of reception; wherein the text stream comprises a plurality of characters; after the number of characters in the text buffer exceeds a specified word count threshold, and when a specified character is received, the characters in the text buffer are used as a sentence, and the audio data of the sentence is generated as a whole. The embodiments of the present application can improve the listening experience of the audio to a certain extent.

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 provides a cross-language information processing method based on digital modeling, and relates to the technical field of data processing, and the method comprises the steps: collecting an ancient language training corpus set with a word number label based on a digital humanity database; pre-training the RoBERTa model to generate a language information processing pre-training model; in order to meet the cross-language requirement, the method comprises the following steps: converting text understanding into images or videos, and constructing a cross-language training data set; finally, a cross-language information processing model is obtained by finely adjusting the pre-training model, and cross-language demand processing is achieved. The method can solve the problems that in the prior art, due to the fact that the performance is poor during ancient language processing, it is difficult to generate images and videos in combination with the specific context of a whole paragraph or a long sentence, the efficiency of executing cross-language tasks is low, the universality and adaptability of a model during processing diversified languages and tasks are further affected, and the processing efficiency is poor. Efficient processing of ancient texts and cross-language tasks is realized, and the understanding and processing efficiency of the model on the ancient texts is remarkably improved.

The application discloses a detection conclusion merging method and device based on a rule engine and a large language model, which comprises the following steps: attaching an original index to each conclusion record in an original conclusion data set; screening out a target rule set containing main merging conditions and subsidiary merging conditions which are effective for a current organization according to an effective range; taking the intersection of the mergeable category labels of the conclusion records and the main merging conditions and the subsidiary merging conditions, and marking as a mergeable main conclusion or a mergeable subsidiary conclusion when the intersection covers all elements of the corresponding merging conditions; taking the mergeable main conclusion as a benchmark to double-traverse and merge into the same merging group, and restoring the order in the group according to the original index; performing field differential aggregation on each merging group; inputting each processing suggestion in the group into a large language model, fusing and outputting a merged conclusion data set under the dynamic word limit. The application solves the technical problems of redundant merging results, field attribute distortion, order disorder and repeated and lengthy suggestion texts.

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 test data set construction method and device for evaluating long text understanding ability of a model, and the method comprises the steps: obtaining a plurality of initial text documents, enabling the document word number of the initial text documents to be larger than a word number threshold value, and / or enabling the time difference between the public time and the current time of the initial text documents to be smaller than a time difference threshold value, and obtaining a plurality of initial text documents; and or the field number of the technical fields to which the plurality of initial text documents belong is greater than a number threshold value; based on the initial text document, constructing processing tasks of different task types; and constructing the test data set based on the processing task and the initial text document. According to the method, the test data set which is larger in text length and has different task types of processing tasks can be constructed, and a foundation is laid for optimizing the language model, so that the language model has the capability of processing super-long text documents.

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.