35results about "Multimedia data clustering/classification" patented technology

Systems and methods for creating a fan profile are disclosed. One embodiment includes querying a user media database. A set of locations are generated based on location metadata associated with media included in the user media database. The embodiment may further include iterating through the user media database to create one or more clusters comprised of similar locations at which one or more media files stored in the user media database have been captured by the user, and filtering out any clusters of the clusters that are less than a threshold size. For each remaining cluster, a time range may be created. All remaining clusters and associated locations within the time range may be categorized and aggregated. All categorized and aggregated remaining clusters may be analyzed to infer demographic traits associated with the user, and a final fan profile associated with the user may be exported based on the analyzing.

This invention discloses a conversational hybrid modal retrieval method, system, device, and medium, comprising the following steps: an intent deconstruction agent decouples unstructured user requests into structured attribute constraints and soft semantic descriptions; a vector enhancement agent uses a generative model to augment the soft semantics with features and fuses visual features through a multimodal cross-attention mechanism to generate enhanced query vectors; a retrieval execution agent performs initial screening using an elastic weighted ranking algorithm combining vector similarity and attribute distance matrices; and a visual auditing agent calls a visual-language model to verify the consistency between the initial screening results and the images, and triggers a closed-loop control module to adaptively adjust parameters and iteratively re-retrieve when the confidence level is below a threshold. This invention improves retrieval accuracy and user intent alignment in long-tail complex query scenarios by constructing a closed-loop mechanism of multimodal semantic enhancement and pre-result self-checking.

The application discloses a content adjustment interaction method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: in response to a task trigger event, displaying task content on a graphical user interface; in response to a content adjustment event for the task content, determining at least one piece of content to be adjusted and an intention to be adjusted corresponding to the content to be adjusted, wherein the content to be adjusted comprises at least one piece of sequence information in the task content; adjusting the content to be adjusted based on the intention to be adjusted, obtaining at least one target content and displaying the target content on the graphical user interface, wherein the content expression of the target content matches the content expression of the content to be adjusted, and the target content can meet the personalized needs of a user when reading the content, thereby improving the learning efficiency of the user for specific content.

Embodiments of the present application provide a media data processing method, device and equipment, and a storage medium. The method comprises: sampling M initial media subsets obtained by an initial clustering model to obtain sampling media data corresponding to the M initial media subsets respectively. According to media description information of the sampling media data corresponding to the M initial media subsets respectively, a predicted theme title of the sampling media data corresponding to the M initial media subsets respectively is predicted by a language model. Clustering accuracy of the initial clustering model for the M initial media subsets respectively is determined according to the predicted theme title of the sampling media data corresponding to the M initial media subsets respectively. The initial clustering model is iteratively trained according to the clustering accuracy of the M initial media subsets respectively until the initial clustering model satisfies a stop iteration training condition, and a target clustering model is obtained. The clustering accuracy of the media data can be improved by the present application.

This invention relates to the field of archival management technology and discloses an intelligent classification and retrieval method for surveying and mapping archives. The method includes: acquiring a user-inputted retrieval command; parsing the retrieval command to obtain terrain constraints; matching a surveying and mapping archive feature vector library based on the terrain constraints to obtain candidate matching results; collecting supporting data based on the user ID and retrieval command; wherein the supporting data includes: scene data, user data, and feature quality data; calculating modal weight parameters based on the terrain constraints, scene data, user data, and feature quality data; constructing a retrieval reward function based on the retrieval command and modal weight parameters; and determining the target retrieval result based on the retrieval command, the retrieval reward function, the surveying and mapping archive feature vector library, and the candidate matching results. By implementing this invention, the problem that related technologies have failed to form a complete surveying and mapping archive retrieval technology system that balances accuracy, real-time performance, and scene adaptability is solved.

The application provides a financial content recommendation method and device, electronic equipment and storage medium, and relates to the field of financial technology or other related technical fields. The method comprises the following steps: in response to an information input operation triggered by a user on a recommendation prompt page, obtaining query information corresponding to the information input operation; obtaining at least one multimedia content matched with the query information and determining label information corresponding to each of the at least one multimedia content; determining target label information matched with personalized information of the user according to the label information corresponding to each of the at least one multimedia content; generating a financial recommendation page according to target multimedia content corresponding to the target label information, and displaying the financial recommendation page. The method improves the accuracy and effectiveness of financial content recommendation.

The application discloses an AI-driven media knowledge base retrieval method based on regional tagging, and particularly relates to the technical field of artificial intelligence, which utilizes a multi-modal large model to perform semantic segmentation on multi-modal media files in a knowledge base, identifies basic units carrying independent semantics, and generates a structured tag set, constructs a cross-modal semantic association network based on the generated structured tag set, wherein nodes represent different media regions, and edges represent semantic similarity, establishes cross-modal links by calculating the semantic similarity of regional tag vectors, vectorizes the current retrieval query of a user, combines historical queries and interaction behaviors to construct a user intent context, performs retrieval in the cross-modal semantic association network, obtains a first candidate set, optimizes the sorting result by using a collaborative reordering algorithm, shows the user retrieval results with high relevance, and dynamically adjusts weights and optimizes the semantic association network according to user interaction.

The application relates to the technical field of search, and discloses a keyword recommendation method, device and equipment and a storage medium, the method comprising the following steps: extracting a search keyword from a search instruction input by a user when the search instruction is received; obtaining a multimedia information keyword candidate set corresponding to the user when the search keyword does not satisfy a preset condition; determining a user interest word set according to a preset fine-grained recognition strategy and the multimedia information keyword candidate set; performing recall on the multimedia information keyword candidate set and the user interest word set respectively according to a preset recall strategy to obtain a plurality of keyword update candidate sets; performing coarse sorting on the keyword update candidate sets; and performing keyword recommendation according to a sorting result. In the application, when a search keyword input by a user does not satisfy a preset condition (i.e. is not standardized or is semantically confused and cannot be recognized), the multimedia information keyword is recommended by fusing a user behavior (i.e. a multimedia information keyword candidate set corresponding to the user) of comprehensive search.

Systems and methods are provided for preventing spoilers in autocompleted search queries. One example method includes, in response to detecting a user interaction at a computing device associated with a search query, identifying a media content history comprising one or more media content items. A predicted search query based on the user interaction and the media content history is generated, and the predicted search query is generated for output, at the computing device.

Embodiments of the present disclosure provide an information presentation method and apparatus, an electronic device and a storage medium. The method includes: presenting a media playing interface in response to a first trigger operation input by a first user, wherein basic media information of a target media is presented in the media playing interface; and in response to a second trigger operation for triggering the presentation of media recommendation information, presenting, in the media playing interface, at least one piece of media recommendation information associated with the target media.

Embodiments of the present application provide a method and device for inquiring, electronic equipment, storage medium and program product, relating to the technical field of artificial intelligence. The method comprises: obtaining case description information of an inquiring object, inputting the case description information into a large inquiring model, determining an involved case type corresponding to the case description information through the large inquiring model, generating corresponding inquiring suggestions based on the case description information and the involved case type through the large inquiring model, and inquiring the inquiring object based on the inquiring suggestions. Through the large inquiring model based on the case description information, the embodiments of the present application classify the case and generate inquiring suggestions, thereby realizing accurate inquiring, obtaining complete and accurate case information, and improving the efficiency of case handling.

The application provides a media data visual analysis system, which comprises a data acquisition module, a data preprocessing module, an analysis engine module, a visual rendering module and a system control module. The data acquisition module is configured with a mixed mechanism of a network crawler and API calling, can adapt to technical protocols of various information sources such as social media, news websites and video platforms, and can continuously and stably obtain original data with different structures. Through the introduction of a heat analysis formula considering time decay and comprehensive behaviors such as likes, comments and shares, the attention degree and propagation effect of information units can be quantified more scientifically and accurately. The application integrates data acquisition, preprocessing, analysis engine and visual rendering into one system, forms a complete closed-loop workflow for media data value mining and display, and overcomes the defects of fragmented links and low integration in the prior art.

An event detection system is configured to access a repository that contains a collection of media content. The media content may for example include images, videos, audio clips, and the like, wherein the media content comprises features that include: tags (e.g., hashtags or other similar mechanisms to label and sort content); captions that comprises one or more words or phrases; continuous numerical values; geolocation data (e.g., geo-hash, check-in data, coordinates); as well as temporal data (e.g., timestamps).

Method of controlling the upload of media content comprising: generating a fingerprint, based on received media content, by applying a trained fingerprint generation neural network to the media content, comprising: an encoder configured to generate an embedding of the media content, the encoder comprising a plurality of encoder layers, each comprising a respective plurality of nodes; and one or more additional layers comprising the embedding, each additional layer of the one or more additional layers comprising a plurality of nodes taking values, wherein the fingerprint comprises a vector comprising the values of a plurality of nodes from the one or more additional layers; accessing a database of reference fingerprints, each reference fingerprint generated based on a respective item of media content; determining whether the generated fingerprint meets a similarity criterion; and generating an alert; and / or preventing upload of the media content.

This invention discloses a method, apparatus, electronic device, and medium for constructing a fire safety knowledge question-and-answer dataset, comprising: acquiring fire safety-related data and segmenting the data into text to obtain fire safety knowledge fragments; labeling the fire safety knowledge fragments based on a prior knowledge framework to obtain multiple types of labels for the fire safety knowledge fragments; extracting semantic features from the fire safety knowledge fragments to obtain semantic feature vectors, and performing nearest neighbor filtering on the semantic feature vectors to obtain a semantic nearest neighbor set; performing professional fusion processing on the semantic nearest neighbor pairs in the semantic nearest neighbor set based on the multiple types of labels corresponding to the fire safety knowledge fragments to obtain the edge weights of the semantic nearest neighbor pairs; clustering the fire safety knowledge fragments based on the edge weights of the semantic nearest neighbor pairs to obtain fire safety knowledge clusters; and constructing a fire safety knowledge question-and-answer dataset based on the fire safety knowledge clusters. The method of this invention deeply integrates professional knowledge in the fire safety field, improving the professionalism and accuracy of fire safety knowledge question-and-answer.

This application provides a method and system for constructing a multimodal large-scale model for safety management in the construction process. The method includes: acquiring structured and unstructured data related to safety management in the construction process, constructing a safety element ontology, and assigning global object identifiers and relational constraints to objects in the structured and unstructured data; processing the structured and unstructured data, generating evidence objects based on global object identifiers and unified time stamps; constructing six categories of libraries based on the evidence objects, using the evidence objects and entries associated with object identifiers in the six categories as training samples, performing multimodal representation alignment training and consistency verification training based on relational constraints to obtain a multimodal large-scale model for safety management in the construction process, and deploying the multimodal large-scale model as a reasoning model for risk classification, hazard closure-loop judgment, and emergency resource allocation based on evidence object references. This application can improve the accuracy of risk identification, accelerate hazard closure-loop handling, and improve emergency response efficiency.

The application discloses a sentiment analysis method based on a multi-modal progressive fusion and enhancement network, which comprises the following steps: firstly, extracting original features of text, audio and video, and obtaining initial sequence features of each mode through time alignment and coding; then, inputting the initial features into a multi-modal progressive fusion module, performing cross-modal interaction and feature fusion in each layer through N-layer stacking, and outputting fusion features after deep interaction; then, sending the fusion features into a multi-modal enhancement module, and obtaining enhanced multi-modal features through multi-layer enhancement optimization; based on the distribution difference between the initial features, calculating the fusion weight between the initial features and the enhanced features, and weighting and integrating the pooled single-modal features and the enhanced fusion features to generate the final multi-modal fusion representation; finally, performing sentiment classification by using the representation, and outputting a classification result. The method effectively integrates multi-modal information and improves the performance of sentiment analysis.

This application relates to the interdisciplinary field of educational information technology and artificial intelligence, and particularly to a method, apparatus, system, and device for classroom information processing. The method includes: acquiring lecture audio streams in a classroom environment using a dedicated acquisition device; processing the acquired lecture audio streams to generate structured classroom notes using knowledge graph technology, wherein the structured classroom notes contain knowledge point information associated with a pre-constructed subject knowledge graph; acquiring images of incorrect questions, identifying and extracting incorrect question information from the images, and associating corresponding knowledge point tags with the incorrect question information based on the pre-constructed subject knowledge graph; and responding to operations on incorrect questions associated with knowledge point tags, locating and presenting the original lecture content related to the knowledge point information in the structured classroom notes based on a reinforcement learning strategy. The technical solution of this application achieves automated and intelligent association of the learning process, improving the accuracy of review.

This disclosure relates to a method, apparatus, electronic device, and storage medium for generating account features. The method includes: acquiring a historical multimedia resource set of an account to be identified; performing feature extraction processing on each historical multimedia resource in the historical multimedia resource set to obtain multiple historical multimedia resource features; and performing feature fusion processing on the multiple historical multimedia resource features to obtain account features of the account to be identified. The account features are used in subsequent processing of associated multimedia resources of the account to be identified. This disclosure extracts multiple historical multimedia resource features from the historical multimedia resource set of the account to be identified and fuses these features to obtain fused features. These fused features can accurately characterize account features. Therefore, using these fused features as account features of the account to be identified improves the accuracy of account features, thereby improving the accuracy of determining resource features of associated multimedia resources.

The application discloses a data processing method and device, a computer readable medium and an electronic device. The method comprises: obtaining an original data set of a target object on a multimedia platform, the original data set being a set of multimedia data; performing clustering processing on the multimedia data in the original data set based on a plurality of preset cluster numbers to obtain clustering results corresponding to the preset cluster numbers; determining a target cluster number according to the clustering results corresponding to the preset cluster numbers, and obtaining a plurality of multimedia clusters according to the clustering results corresponding to the target cluster number; and screening the plurality of multimedia clusters according to the multimedia data aggregation degree corresponding to each multimedia cluster to obtain a target multimedia cluster of the target object. The multimedia data in the target multimedia cluster obtained by the application is more consistent with the interest points of the target object, thereby improving the judgment accuracy of the interest points of the target object.

The application discloses a personalized development guidance and resource matching system based on a student digital portrait and relates to the technical field of intelligent education. The system comprises a student digital portrait construction module for constructing a dynamic state vector of a student in a multi-dimensional feature space; wherein the multi-dimensional feature space at least contains an ability dimension, an interest dimension and a learning style dimension; and an educational resource semantic graph module for constructing resource feature vectors of educational resources in the multi-dimensional feature space and a semantic association network among the resources. The application establishes a dynamic balance mechanism of individual development efficiency and group fairness, fundamentally solves the adaptability trap and stereotype reinforcement problem caused by the traditional recommendation system taking the maximum matching degree as a single target, ensures that students from different backgrounds can obtain personalized paths considering the current foundation and development potential, and makes the algorithm a real promoter of educational fairness rather than a replicator of bias.

The application belongs to the technical field of multimedia, and particularly relates to a multimedia content pushing optimization method and system based on artificial intelligence, which comprises the following steps: determining user characteristics based on user historical behavior information; comprising: obtaining historical browsing records, playing history, search history, collection records, like history and comment history of the user to combine as user historical behavior information; extracting content browsing preferences of the user from the user historical behavior information. The application determines the user characteristics by constructing a mapping from the user historical behavior information to the key behavior node and screening the preliminary label structure based on a preset evaluation score, converts the user behavior into a hierarchical clear and quantifiable feature model, overcomes the problem of insufficient recognition accuracy caused by the dependence of the prior art on simple label aggregation, and deeply mines the potential interest of the user to lay a foundation for subsequent accurate pushing.

A method and system for vector retrieval acceleration includes at least one calculation module and at least one sorting acceleration module. Each calculation module includes a first receiving end and a second receiving end for receiving query vector blocks and center vector blocks respectively, and performs operations on the received query vector blocks and center vector blocks, so as to obtain distance calculation results between each query vector and all center vectors, sorting acceleration module. The at least one calculation module corresponds one-to-one with the at least one sorting acceleration module. Each sorting acceleration module receives the distance calculation results output by the corresponding calculation module, sorts the distance calculation results, so as to obtain a sorting intermediate state, and then stores sorting intermediate states into a system memory.

This invention relates to a method and system for generating three-dimensional objects based on semantic analysis. More specifically, this invention relates to a method and system for analyzing text-based documents and automatically generating three-dimensional objects corresponding to the meaning of those documents.

The application discloses a public safety command intelligent processing method based on multi-modal fusion deep learning, belongs to the technical field of public safety command, and aims to solve the problems of unclear field perception, delayed abnormal early warning and insufficient joint linkage of the existing public safety command. The metadata model supporting mode adaptation and relationship customization is constructed, multi-modal and billion-level data convergence fusion is realized, cross-domain and cross-layer streaming media processing is cooperated with video content analysis and mining, the smoothness of media transmission is ensured, the accuracy of abnormal event time section and space section positioning in monitoring video is improved, and the identification accuracy under complex conditions is ensured. Based on the multi-modal information fusion of expert knowledge model and deep learning model in the decision-making stage, the research and judgment accuracy is greatly improved, and the practical ability of the professional research and judgment model is supported. The application is suitable for the public safety command intelligent processing method based on multi-modal fusion deep learning.

This invention discloses a multimodal retrieval method combining image features and semantic understanding, belonging to the field of information retrieval technology. This application employs a design that combines image segmentation technology with semantic role labeling, fully mining the semantic information in image and text data. By generating semantically anchored visual codebooks and cross-modal anchored indexes, deep fusion of images and text is achieved. This method can more accurately correspond and align visual and semantic information in images and text, alleviating the problems of information loss and low retrieval accuracy caused by independent processing of image and text data in existing technologies, significantly improving the accuracy and efficiency of multimodal retrieval. Simultaneously, by using anchor vectors from both the image and text sides for indexing, visual codewords and semantic role slots are tightly integrated, establishing more accurate cross-modal matching for matching and indexing between different modalities.

This invention discloses a method and system for intelligent generation and processing of meeting content based on a multimodal large model, comprising: collecting raw meeting data and performing standardized preprocessing; inputting the multimodal large model, extracting multimodal features and semantically aligning them; performing cross-modal hash encoding to generate binary codes and establish an index library; hash-retrieval of relevant segments to construct a directed graph of meeting content; optimizing the path using Monte Carlo tree search based on the directed graph structure; generating structured meeting content and outputting minutes, summaries, and action items. This invention achieves efficient extraction, accurate retrieval, and intelligent structured generation of meeting content by integrating a multimodal large model, cross-modal hash encoding, and Monte Carlo tree search.

On the basis of the embodiments of the present disclosure, provided are a method and apparatus for determining event information, and a device and a storage medium. The method for determining event information comprises: classifying acquired media content according to a plurality of predetermined categories, wherein the plurality of predetermined categories respectively correspond to different degrees of event association; in response to the media content being classified into a target category among the plurality of predetermined categories, on the basis of the media content, determining event information of a target event in a plurality of dimensions; and storing the event information in the plurality of dimensions in an associated manner, so as to represent the target event. Therefore, a target event can be represented from a plurality of different dimensions. In this way, in application scenarios such as push and query, recommendations can be provided to a user on the basis of event information in different dimensions, thereby improving the user experience.

The application discloses a kind of teaching archives digitization automatic indexing and multidimensional association analysis system, it is related to education information technology field, including: data acquisition module gathers multiple-source heterogeneous teaching archives data and is converted into unified format data to be handled;Intelligent indexing module is according to teaching field special glossary to the semantic analysis and multimodal adaptation of data to be handled, generates multidimensional indexing information;Multidimensional association analysis module identifies direct association according to multidimensional indexing information, excavates implicit association and carries out time series evolution analysis, reveals internal law of teaching;Knowledge graph construction module extracts teaching entity, establishes entity relationship based on association analysis result, constructs dynamic updating knowledge graph;Application service module supports retrieval based on knowledge graph, personalized push and visual analysis, and sets fine-grained access permission.The application can improve teaching resource management efficiency and utilization value, guarantee data security and support teaching decision optimization.