36 results about "Topic analysis" patented technology

The invention provides a social listening data noise reduction method and device based on a large language model, and belongs to the technical field of social media data processing. The method comprises the steps that social media data are collected and preliminarily filtered to obtain preliminarily filtered data; generating a semantic embedding vector by using a pre-trained large language model, and screening semantic related data in combination with cosine similarity and context; identifying and filtering duplicate and variant contents through an embedded model and a generative adversarial network to obtain duplicate-removed data; performing multi-modal feature extraction on the de-duplicated data, and screening credible data in combination with generative AI detection and sentiment analysis; and filtering semantic inconsistency and non-topic continuation texts by adopting a clustering algorithm and semantic relationship analysis to finish noise reduction. According to the method, through a multi-stage and multi-dimensional noise reduction strategy, the purity and credibility of social listening data are effectively improved, and high-quality data support is provided for subsequent topic analysis and decision making.

The present invention discloses a knowledge point extraction method and system based on a large model, which comprises the following steps: obtaining a knowledge text to be extracted, and obtaining important words of the text through large model preprocessing; vectorizing the important words of the text obtained through preprocessing to obtain text word vectors, and obtaining a text vector data set through aggregation; clustering analysis of the text vector data set to divide the text vector data set into different clusters; analyzing the text word vectors, selecting keywords representing the theme of the cluster from each cluster as the knowledge theme of the cluster; analyzing the degree of association between each important word of the text and the knowledge theme in each cluster, and extracting the important words of the text that meet preset conditions as the knowledge points of each cluster; and continuously aggregating and updating the knowledge points extracted from each cluster to obtain a final knowledge point set. The present invention can extract useful knowledge information from text data, and help better understand the content and theme of the text data by continuously updating and improving the knowledge point set.

This invention relates to a big data topic analysis method based on an embedding model. First, the Sentence-BERT model is used to perform sentence embedding representation on preprocessed Chinese text data. Then, the UMAP projection dimensionality reduction algorithm is used to reduce the dimensionality of the embedded vectors. Next, the HDBSCAN clustering algorithm is used to cluster the dimensionality-reduced vectors. Based on the assignment of each Chinese text in the target Chinese dataset to a corresponding topic class, the Chinese words with the highest c-TF-IDF scores are selected to represent each topic class. Finally, the DSG model is used to perform word embedding representation on the topic words, calculating the similarity between different topic words and between different topic classes, thereby detecting the volatility of newly emerging topic classes. The entire scheme design has higher topic consistency and topic diversity, and can detect new hot topics in a timely and accurate manner, providing early warnings.

This application relates to a method, apparatus, computer device, and storage medium for detecting hidden encryption attacks. The method includes: performing topic analysis on input text content to determine the topic semantics of the text content; performing intent parsing on the text content to extract multiple intent information contained in the text content and obtain the intent semantics corresponding to each intent information; performing anomaly analysis based on multiple intent semantics to determine abnormal intent information; determining the semantic similarity between the intent semantics of the abnormal intent information and the topic semantics, and identifying abnormal intent information with a semantic similarity lower than a preset similarity threshold as irrelevant intent information; and performing instruction nature identification on the irrelevant intent information to determine the hidden encryption attack detection result of the text content. Using this application can improve the ability to identify malicious content with high concealment and semantic complexity.

The present invention discloses a keyword-related topic analysis method based on local word embedding technology. When a user enters a keyword, the method uses global word embedding technology to analyze the keyword in text big data, sorts the cosine similarity between the vocabulary and the keyword in the data, and selects the subject words. The subject dictionary is constructed based on the subject words, and the text big database is segmented by time, author, article type, etc. The local word embedding analysis is performed on the keywords and the subject dictionary within the segment, and the local cosine similarity in each segment is used to obtain the theme of each segment. Then, the theme change of the keyword in the text big database can be analyzed. The present invention can be widely used in the fields of social science, public opinion monitoring, historical document analysis, etc., and provides an efficient and accurate theme dynamic analysis tool for text mining.

The application provides a self-adaptive LDA topic model training system based on public opinion real-time data flow, comprising a data gathering module, a data preprocessing module, a self-adaptive LDA model training module and an incremental LDA model fusion module; the data gathering module is used for extracting and converting and loading the structured and semi-structured data and inputting into a distributed message bus kafka; the data preprocessing module is used for preprocessing the data in the message bus kafka and finally forming a weighted word vector; the self-adaptive LDA model training module is used for training to obtain an LDA model result and merging the training result; the incremental LDA model fusion module is used for fusion training to generate a new round of LDA model. The application is superior to the traditional LDA topic analysis method in accuracy and performance, and is applied to network public opinion field event detection, recommendation, word cloud and retrieval and other practical engineering projects, and creates commercial value.

The application relates to a customer complaint text classification method and device, computer equipment and a storage medium. The method comprises the following steps: preprocessing text data to be processed to obtain a plurality of word segmentation results; performing vectorization processing on the plurality of word segmentation results to obtain first vector values of the word segmentation results; inputting the plurality of word segmentation results into a pre-trained topic analysis model to obtain second vector values of each customer complaint topic to which the text data to be processed belongs and third vector values of keywords of the text data to be processed; performing splicing processing on the first vector values, the second vector values and the third vector values to obtain splicing features, and taking the splicing features as features of the text data to be processed; and inputting the features of the text data to be processed into a topic classification model and a perception classification model of a pre-trained classification model respectively to obtain a customer complaint topic category and a customer complaint perception category to which the text data to be processed belongs. The method can deeply analyze specific customer complaint contents under a field category.

The present invention relates to a method and a system utilizing a large language model for topic allocation based on quotations. According to the present invention, a computer-implemented method utilizing a large language model for topic allocation based on quotations may comprise the steps of: specifying content to be analyzed, which includes at least one sentence; specifying at least one topic related to the at least one sentence included in the content to be analyzed; extracting, by using a large language model, at least one quotation corresponding to each of the at least one topic from the content to be analyzed; and providing, by using the topic and the quotation, a topic analysis result for the content to be analyzed.

The invention relates to the field of natural language processing, and discloses a Chinese language and literature text topic analysis system based on big data, and the system comprises a core semantic primitive library construction module which is used for extracting and constructing a knowledge base containing semantics, culture and timestamps from a meta-poeia corpus offline; the text multi-dimensional analysis module is used for extracting four-dimensional features of grammar, core semantics, intertext and emotion time sequence of the to-be-analyzed text in parallel; the feature collaborative analysis module is used for fusing multi-dimensional features, performing context dynamic activation on primitives, and innovatively quantifying and detecting deep semantic conflicts by calculating the dispersion of feature vectors in a semantic space; and the topic report generation module is used for synthesizing a macroscopic topic, integrating all analysis results and generating a structured multi-level report. The system can objectively and deeply reveal complex and even contradictory theme connotation in literature works through modular cooperation, and the accuracy and depth of text analysis are remarkably improved.

The invention discloses an NLP-based method for realizing automatic transcription of hot word extraction specifications, which comprises the following steps of: presetting different scenes, defining hot word types in the scenes and constructing a common hot word library, and preprocessing and cleaning a document; automatically analyzing the cleaned context data, and automatically extracting hot words in the corresponding file content; the extracted hot words are intelligently classified, and a dynamic hot word library is constructed; meanwhile, a voice transfer result is transferred into a hot word verification module in real time; carrying out semantic analysis according to the context of the voice recognition result, judging the most coincident hot word, and carrying out replacement display on the extracted hot word result; according to the method, a normalized result of hot word replacement is displayed, a text result of speech recognition is automatically adjusted, weights of hot words in a dynamic hot word library are updated in real time, representative hot words are extracted from preprocessed texts based on a trained NLP semantic model by applying technologies such as a clustering algorithm and topic analysis, and the accuracy of transcription is improved.

The invention provides a method for automatically generating and distributing network guidance content based on a large model, which comprises the following steps of: 1, data monitoring: collecting related data of a monitoring target in the Internet; 2, analyzing a hot topic, processing the related data collected in the step 1, and obtaining the topic of the hot event and the corresponding data content; step 3, guide content generation: setting a caliber requirement of the guide content according to the theme of the hot event obtained in the step 2, and generating the guide content through a large language model; and 4, task distribution management: selecting a distribution method and a distribution strategy according to the guide content obtained in the step 3, and carrying out guide content distribution.

The application relates to an open domain long text classification method and device based on theme analysis. The method comprises the following steps: constructing a field self-adaptive word table; preprocessing original long text to obtain purified text; using an LDA model to construct a latent theme space; extracting a core sentence group from the long text through semantic clustering to generate an abstract; mapping the abstract to the theme space to calculate a correlation degree score and output a theme identification; and converting the theme identification into a business classification label by querying a theme-label mapping table. The application effectively solves the problems of complex long text semantics and dynamic label system changes, and improves classification accuracy and expansibility.

The invention relates to the technical field of medical big data processing, in particular to a neurological critical patient follow-up visit management method and system. The method comprises the following steps: acquiring complaint data of a patient; quantitatively evaluating the health status of the patient by using a questionnaire to obtain a first score; if the first score is greater than a questionnaire threshold value, performing topic analysis on the chief complaint data to obtain a first topic; performing topic analysis on a question corresponding to the chief complaint data to obtain a second topic; calculating the semantic similarity between the first theme and the second theme; constructing a definition vector based on the semantic similarity and the first score; calculating a first similarity between the definition vector and a preset standard vector; when the first similarity meets the first range, risk assessment is carried out based on the medical information, the medical report and the chief complaint data of the patient, intervention measures of the patient are determined according to an assessment result, the recovery state of the neurological function of the patient can be continuously tracked, and medical staff are helped to comprehensively know the rehabilitation process of the patient.

The present application provides an interest identification method, apparatus, device and storage medium, which relates to the field of big data. The method includes: obtaining voice call data, converting the voice call data into text data; constructing a heterogeneous graph based on the text data, and determining the high-frequency words in the heterogeneous graph by statistical methods; classifying the heterogeneous graph by means of a graph neural network to obtain at least one category corresponding to the text data; performing topic analysis on the documents in each category by means of latent Dirichlet distribution to obtain the topic of each category and the keywords in each topic; determining user interests based on the keywords and high-frequency words in each topic. The method of the present application realizes the rapid and accurate identification of user interests, avoids the inaccuracy of manual identification of user interests and the corresponding waste of personnel, reduces the waste of manpower, reduces the time of user interest analysis, and improves the efficiency of user interest identification.

This invention discloses a method and apparatus for predicting the emergent behavior of groups based on social network evolution computation. By deeply analyzing the evolutionary laws and group behavior characteristics of social networks, and comprehensively considering the evolutionary laws, group behavior characteristics, and interactions between individuals, this invention can simultaneously simulate the evolutionary process of social networks and changes in group behavior from two dimensions: network structure and node properties. This overcomes the limitations of traditional methods that only analyze from a single dimension, thus significantly improving the accuracy and reliability of group behavior emergent prediction. Furthermore, this invention can accurately identify key nodes, information propagation paths, and driving factors of group behavior. By combining pre-trained models and sentiment dictionaries to analyze text data, it improves the accuracy of sentiment recognition and topic analysis, thereby enhancing the targeting and precision of group behavior prediction.

This application discloses an article classification method, apparatus, electronic device, and storage medium, relating to the field of artificial intelligence technology or other related fields. The method includes: acquiring at least one target article to be classified; performing topic analysis on the target article using a preset model to generate topic tags; calling a classification tag document and querying the domain to which the topic tags belong in the classification tag document to obtain the domain tags of the target article; querying all topics involved by the topic tags and domain tags in the classification tag document to obtain a candidate topic set, and combining the preset model and the candidate topic set to perform topic analysis on the target article to generate topic tags; summarizing the domain tags, topic tags, and topic tags to generate the article classification result for the target article. This application solves the technical problem in related technologies where reliance on manual labeling and fixed classification rules leads to low accuracy in article classification.

The embodiment of the invention provides a comment information publishing method and device and a storage medium. A workflow engine integrated in an AI agent is used in combination with an MCP tool specially developed for calling a large-scale language model, and a workflow for automatic comment is defined in the workflow engine; the AI agent calls the MCP tool by depending on a large language model to complete the execution of the workflow; specifically, in response to a comment trigger event, target scene information of a post to be commented is determined; and generating model cue words based on the dependency relationship of the workflow and the MCP tool, inputting the model cue words into the large language model, and guiding the model cue words to execute the steps of post retrieval, post acquisition, topic analysis, comment generation and comment release included in the workflow. And the corresponding MCP tool is called by the large language model in the step of having the dependency relationship with the MCP tool, and the comment information is automatically generated and published to the service platform, so that the efficiency and the accuracy of publishing the comment information are improved, and the customer satisfaction is further improved.

The invention relates to a deep research method and device of a script, a storage medium and electronic equipment. The method comprises the steps of performing structured analysis on a to-be-deeply researched script text to extract script analysis data including role information, session information and candidate themes, and generating a script theme set according to the script analysis data; generating at least one corresponding target audience portrait for each script theme in the script theme set; retrieving and analyzing social public opinion data associated with each target audience portrait to obtain public opinion analysis result data, and performing matching analysis on the public opinion analysis result data and the script theme set to obtain topic analysis result data; and performing aggregation and consistency verification processing on the public opinion analysis result data and the topic analysis result data to obtain a structured deep research report of the script text. According to the method and the device, the technical problem of lack of iterable, multi-dimensional linkage and verifiable script deep analysis is solved.

The invention discloses a data collaborative analysis method based on a unified index view, and relates to the technical field of data governance and intelligent analysis. Long-term pain points of disordered index definition, inconsistent analysis calibers and inefficient cross-department cooperation in an enterprise multi-source heterogeneous environment are effectively solved, the credibility of data assets is improved, a unified index model solidifies calculation rules and statistical calibers from the source, core indexes such as sales and user retention rates are enabled to be semantically consistent in the whole platform, and the system performance is improved. Secondly, the collaborative analysis efficiency is released, business personnel do not need to repeatedly confirm the meaning of the index, a cross-department comparable report can be quickly constructed based on a preset theme analysis view, meanwhile, a technical team greatly reduces the operation and maintenance cost through log binding and change tracing, in addition, the technical adaptability can be enhanced, and the method is suitable for a multi-source heterogeneous data environment. The platform flexibly responds to business scene changes through a dynamic mapping strategy, and the problem of stiffness of the index tree is avoided.

This invention discloses a visualization analysis method, apparatus, device, and storage medium for multi-indicator comparison. The method includes: in response to a user's topic configuration operation in the visualization analysis interface, creating a target analysis topic and binding a target analysis chart to the target analysis topic; in response to a user's indicator selection operation in the visualization analysis interface, selecting at least one target indicator from a unified indicator library, aligning each target indicator with the standard dimensions of the target analysis chart, and then displaying them superimposed in the same coordinate system; wherein the indicators in the unified indicator library are all constructed based on predefined indicator standard models; and based on the target indicators superimposed in the same coordinate system, performing analysis using a preset multi-level indicator comparison strategy to obtain multi-level analysis results corresponding to the target analysis topic. Using this invention can effectively shorten the topic analysis cycle and improve topic analysis efficiency.

The application belongs to the field of social network derived topic analysis, and particularly relates to a key influential user discovery method for time-varying network, which comprises the following steps: preprocessing original topic network data to obtain topic network data; slicing the topic network data along the time dimension to obtain topic network data at different time points; calculating the dynamic importance of user nodes at different time points, and obtaining the transfer sequence of each user node according to the dynamic importance; constructing the feature vector of each user node by using an attention mechanism; obtaining the cumulative interaction sequence and the recent interaction sequence of each user node, and obtaining the interest vector of each user node through a double-Transformer structure; splicing the feature vector and the interest vector of each user node to obtain comprehensive features; inputting all the comprehensive features into a full connection layer and a softmax function layer to obtain the key influential user identification result; and the application can effectively discover key influential elements in the topic space.

The invention discloses a voice interaction method, device and equipment, and the method comprises the steps: recognizing the voice of a current interaction request, and obtaining the voice of each user; converting the voice of each user into a text; performing semantic analysis on the text corresponding to the voice of each user, extracting keywords, and performing topic analysis on the text corresponding to the voice of each user to obtain topics corresponding to the text; and calling a large language model, generating a personalized reply based on the text, the keyword, the information stored in the two-dimensional storage structure and the information stored in the global structure of the user in the current round of interaction, and converting the generated personalized reply into voice to be output. According to the method, parallel management and response to multiple topics are realized, the problems of semantic loss and context misjudgment are effectively solved, and the practicability and user experience of intelligent equipment in a complex social scene are enhanced.

Systems and methods for providing enhanced teleconferencing. An example method includes receiving audio streams from a plurality of client devices of participants of a teleconference; converting the audio streams for a first conversation within the teleconference into first text; converting the audio streams for a second conversation within the teleconference into a second text; analyzing the first text to identify one or more topics being discussed in the first conversation; analyzing the second text to identify one or more topics being discussed in the second conversation; and presenting, in a teleconference user interface, at least one of the one or more topics being discussed in the first conversation or the one or more topics being discussed in the second conversation.

The invention discloses an unreasonable medical examination judgment method, system and device and a storage medium, and the method comprises the steps: firstly building an observation list of examination items through a statistical analysis and cross verification method, and then carrying out the analysis and processing of patient samples related to the examination items in the observation list; performing topic analysis on a patient sample through an LDA algorithm to obtain a disease topic, calculating the correlation between each disease and the mode disease, and performing cross validation on the disease which is remarkably different from the mode disease through the disease topic to obtain a disease result; and judging whether the examination sheet is unreasonable or not under the condition of the illness, so that the system can perform unreasonable examination, monitoring and early warning on the condition. According to the method, the importance of examination items is considered, judgment of unreasonable examination is carried out on the basis of objective illness condition facts and probability distribution in the examination sheet making process, reasonability and basis are achieved, and the problem that at present, unreasonable examination is difficult to judge is effectively solved.

A computer implemented method includes obtaining, by one or more processors, a learning objective for a class session. The student interactions with computing devices are monitored during the class session to collect student interaction data. The collected student interaction data is analyzed using a machine learning model trained to identify patterns indicative of engagement or disengagement with the learning objective, wherein the machine learning model applies a topic analysis algorithm to content accessed by the students to determine a relevance score relative to the predefined learning objective. An engagement status is determined for each student based on the relevance score and a predetermined engagement threshold. Real-time feedback is generated for an educator based on the engagement status or each student, wherein the feedback includes actionable recommendations for interventions to enhance engagement.

The application relates to a topic analysis method and device, computer equipment and a storage medium. The method comprises the following steps: extracting a plurality of keywords from conversation data; screening topic keywords from the plurality of keywords; determining associated keywords related to the topic keywords from the plurality of keywords; combining the topic keywords and the corresponding associated keywords respectively to obtain at least one topic phrase; and determining a target topic corresponding to the conversation data based on the at least one topic phrase. The method can improve the efficiency of topic analysis.

The invention relates to an open domain long text classification method and device based on topic analysis. Comprising the following steps: constructing a field adaptive word list; preprocessing the original long text to obtain a purified text; constructing a potential theme space by using an LDA model; extracting a core sentence group from the long text through semantic clustering to generate an abstract; mapping the abstract to a theme space, calculating a correlation score and outputting a theme identifier; and converting the theme identifier into a service classification tag by querying the theme-tag mapping table. According to the method, the problems of long text semantic complexity and label system dynamic change are effectively solved, and the classification accuracy and expansibility are improved.

Disclosed herein are systems and methods for generating a prompt with context based on a list of topics generated for responses from large language models (LLMs). In one aspect, the method includes: obtaining a query from a user; generating and transmitting a prompt based on the query for input into a first and second LLMs; obtaining a first selected portion of the first response from the first LLM and at least a second selected portion of the second response from the second LLM; generating a list of topics using a trained topic analysis machine learning model (MLM) to identify topics from the responses; and generating a prompt for input into the third LLM utilizing at least one topic from the list of topics using at least the selected portions from the first LLM or the selected portions from the second LLM.

The invention belongs to the technical field of travel route recommendation, and particularly discloses a travel route recommendation method based on LLM user portraits and multi-dimensional feature optimization, and the method comprises the steps: calling LLM to analyze explicit and implicit preferences to construct user portraits based on travel demand texts and historical search records; performing topic analysis based on the unstructured text data to determine a topic probability, and generating a static feature vector in combination with the structured information; calculating the matching degree of the portrait and the static vector, and screening candidate points; on the basis of the candidate point coordinates, the travel time consumption and the real-time traffic, path search travel serialization is carried out with the purposes of minimizing the passing time length and maximizing the experience satisfaction degree, and a travel path plan is generated; structured information, BERT emotion and an LDA theme are fused to construct a static feature vector, an entropy weight method is combined with a subjective weight to determine a comprehensive weight, multi-objective optimization is performed through a path planning algorithm accessing real-time traffic information, and personalized recommendation of tourism paths is realized.

The disclosed embodiments provide techniques for generating scripts for TV / Web series using viewer feedback. The proposed system captures free form user input data from social media websites and streaming media platforms, including viewer comments and ratings. System uses advance techniques such as natural language processing (NLP) for sentiment and topic analysis and generative AI for content generation. A feedback loop with professional scriptwriters ensures the generated scripts are refined, polished and finalized with human oversight. Traditional methods of gathering viewer insights, such as ratings and surveys, often lack the ability to capture free form audience opinions and utilize the same to influence the future content development. This AI-driven approach significantly reduces the time and effort required by traditional scriptwriting methods while aligning entertainment content with audience preferences, fostering deeper audience involvement and satisfaction.