7 results about "Anchor text" patented technology

The invention relates to a device and a method for converting anchor voice to generate live streaming in different languages, and a computer readable medium. The method comprises the following steps of: detecting an anchor language type in an anchor voice streaming and converting the anchor voice streaming into anchor text information belonging to the anchor language type; the method comprises the following steps: translating anchor text information into target text information corresponding to a target language, converting the target text information into a target voice streaming according to streaming time information of the target text information, and synthesizing a live video streaming and a target voice streaming. According to the invention, the multilingual live streaming with synchronously corresponding translation content and correct translation can be generated, and the technical effect of improving the translation semantic consistency and generation efficiency of the multilingual streaming is achieved.

Technology is disclosed for generating a comprehensive or large index. In an example embodiment, the Internet is crawled to collect URLs corresponding to webpage addresses, and infrequent or rare terms are extracted from the URL, anchor text, or title of a webpage. An index is populated that maps these rare terms to URLs, enabling fast lookups during search queries. In some instances, common terms are also extracted and used to populate an index associating common terms with URLs. When a query is received, the comprehensive or large index is used to retrieve URLs relevant to both niche and mainstream content causing a balanced integration of specialized and popular results within the search output.

The invention discloses an AI broadcasting method, an AI broadcaster generation system, electronic equipment and a medium, and relates to the technical field of artificial intelligence. In the method, an anchor report text and an interactive text are converted into audios respectively; according to the method, the AI anchors are broadcasted according to the initial broadcasting time of the AI anchors, and then the audios of the AI anchors are broadcasted according to the initial broadcasting time of the audios, so that cooperative conversation among the multiple AI anchors is realized, a real broadcasting scene is met, and the orderliness and rationality of broadcasting of the multiple AI anchors are ensured; secondly, when information input by the user is detected, current broadcasting is stopped, a reply text is generated to respond to the user, and real-time interaction between the AI anchor and the user is achieved; and thirdly, calling a pre-training language model, generating a reply text in combination with the broadcaster text, the broadcast interruption position and the information input by the user, and responding to the user by using the reply text, so that the response of the AI anchor is related to the information input by the user, the AI anchor can accurately return to the original principal line after the interaction is finished, the core theme is ensured not to be deviated, and the user experience is improved. And the overall structure integrity of the broadcast is maintained.

Technology is disclosed for generating a comprehensive or large index. In an example embodiment, the Internet is crawled to collect URLs corresponding to webpage addresses, and infrequent or rare terms are extracted from the URL, anchor text, or title of a webpage. An index is populated that maps these rare terms to URLs, enabling fast lookups during search queries. In some instances, common terms are also extracted and used to populate an index associating common terms with URLs. When a query is received, the comprehensive or large index is used to retrieve URLs relevant to both niche and mainstream content causing a balanced integration of specialized and popular results within the search output.

The invention provides a table-based multi-modal retrieval enhancement generation method, which comprises the following steps of: S1, obtaining a sub-query or a sub-query of a user, screening out a table with the highest correlation with the sub-query from a table library according to the sub-query, and serializing the table to obtain an anchor point table set; s2, screening out an anchor point image set from an image library according to the anchor point table set, and screening out an anchor point text set from a text library according to the anchor point table set; s3, by taking the anchor point table set as a center, constructing modal association boundaries among the anchor point table set, the anchor point image set and the anchor point text set to obtain a multi-modal knowledge graph; s4, the weight of the center position of the multi-modal knowledge graph is amplified, a retrieval result is output, a query generator is established through the adjusted multi-modal knowledge graph, the query generator can generate a new round of query, whether the user terminates the retrieval or not is judged, and if yes, the process is ended; and if not, re-executing the step S1 by taking a new round of query as input.

The application provides a method for quickly constructing overlapping relationship extraction text data, comprising the following steps: step 1, constructing a field dictionary; step 2, obtaining text; step 3, dividing sentences and assembling data; step 4, adding anchor text to a candidate entity set; step 5, extracting non-anchor text entities in the text, and adjusting the data distribution of samples; step 6, inquiring the entity number of an entity in a Wikipedia knowledge graph, which is used as the unique identifier of the entity; and step 7, inquiring the relationship between entities, and marking the relationship of the divided sentences. The application redefines the type of overlapping relationship data, provides a data set that can be studied, and lays a data foundation for overlapping relationship research.

The invention relates to a method for predicting an illegal parking motivation of a truck, and the method comprises the steps: obtaining a related judgment document of illegal parking of the truck as a data source, screening out an effective sample through preprocessing, carrying out the manual marking of the effective sample according to a defined illegal parking motivation type, and dividing the effective sample into a training set and a test set; based on the processed data set, constructing a three-dimensional feature system of the traditional text, the depth semantics and the anchor text, respectively extracting features of the traditional text, features of the depth semantics and features of the anchor text, splicing and fusing the three extracted features, and obtaining a multi-dimensional feature vector after dimension reduction and noise filtering; and on the basis of the obtained multi-dimensional feature vectors, an integrated learning framework is established, a multi-classification strategy is adopted to train exclusive binary classifiers for various illegal parking incentives, prediction results of the classifiers are aggregated through a voting mechanism, and prediction of the illegal parking incentives of the trucks is realized. Compared with the prior art, the method has the advantages of realizing accurate and robust prediction of the illegal parking motivation of the truck and the like.