Abstract: A large model optimization training method in the artificial intelligence fields, such as large models, deep learning, natural language processing, may include: taking, as candidate queries, queries collected from a predetermined data source and capable of serving as input to a large model in response to determining that an optimization triggering condition is met; screening out target queries from the candidate queries, the target queries being queries which cannot be correctly processed by the large model; and constructing respectively corresponding training samples according to the target queries, the training samples being used for carrying out optimization training on the large model.

Abstract: A video processing method, an electronic device and a storage medium are provided, and relate to the field of artificial intelligence, and particularly relates to the fields of deep learning, model training, knowledge mapping, video processing and the like. The method includes: acquiring a plurality of first video frames, and performing fine-grained splitting on the plurality of first video frames to obtain a plurality of second video frames; performing feature encoding on the plurality of second video frames according to multi-mode information related to the plurality of second video frames, to obtain feature fusion information for characterizing fusion of the multi-mode information; and performing similarity matching on the plurality of second video frames according to the feature fusion information, and obtaining a target video according to a result of the similarity matching.

Abstract: A technical solution for processing a model generation result, which relates to the field of artificial intelligence technologies is disclosed. An implementation includes: disassembling a text generation result of a generative large model to obtain a plurality of result logic units; wherein each result logic unit includes a segment in the text generation result; each segment is capable of independently identifying one premise or conclusion in a logical inference relationship of the text generation result; and the text generation result is a response result generated by the generative large model based on text input information; generating a logical inference graph capable of characterizing a logical inference relationship among the plurality of result logic units based on the plurality of result logic units; and determining whether logical inference of generation of the text generation result by the generative large model is correct or not based on the logical inference graph.

Abstract: A theme classification method based on multimodality is related to a field of a knowledge map. The method includes obtaining text information and non-text information of an object to be classified. The non-text information includes at least one of visual information and audio information. The method also includes determining an entity set of the text information based on a pre-established knowledge base, and then extracting a text feature of the object based on the text information and the entity set. The method also includes determining a theme classification of the object based on the text feature and a non-text feature of the object.

Abstract: A method, electronic device and storage medium for generating information are disclosed. The method includes: acquiring a plurality of tag entity words from a target video, the tag entity words including a person entity word, a work entity word, a video category entity word, and a video core entity word, the video core entity word including an entity word for characterizing a content related to the target video; linking, for a tag entity word among the plurality of tag entity words, the tag entity word to a node of a preset knowledge graph; determining semantic information of the target video based on a linking result of each of the tag entity words; and structuring the semantic information of the target video based on a relationship between the node and an edge of the knowledge graph, to obtain structured semantic information of the target video.

Abstract: Embodiments of the disclosure disclose a method, apparatus, server, and storage medium for incorporating a structured entity, wherein the method for incorporating a structured entity can comprise: selecting a candidate entity associated with a to-be-incorporated structured entity from a knowledge graph, determining the to-be-incorporated structured entity being an associated entity based on prior attribute information of a category of the candidate entity and a preset model, merging the associated entity and the candidate entity, and incorporating the associated entity into the knowledge graph. The embodiments can select a candidate entity, and then integrate a preset model using prior knowledge, which can effectively improve the efficiency and accuracy in associating entities, and reduce the amount of calculation, to enable the structured entity to be simply and efficiently incorporated into the knowledge graph.

Abstract: A method for mining an entity focus in a text may include: performing word and phrase feature extraction on an input text; inputting an extracted word and phrase feature into a text coding network for coding, to obtain a coding sequence of the input text; processing the coding sequence of the input text using a core entity labeling network to predict a position of a core entity in the input text; extracting a subsequence corresponding to the core entity in the input text from the coding sequence of the input text, based on the position of the core entity in the input text; and predicting a position of a focus corresponding to the core entity in the input text using a focus labeling network, based on the coding sequence of the input text and the subsequence corresponding to the core entity in the input text.

Abstract: A method, apparatus, device, and storage medium for linking an entity, relates to the technical fields of knowledge graph and deep learning are provided. The method may include: acquiring a target text; determining at least one entity mention included in the target text and a candidate entity corresponding to each entity mention; determining an embedding vector of each candidate entity based on the each candidate entity and a preset entity embedding vector determination model; determining context semantic information of the target text based on the target text and each embedding vector; determining type information of the at least one entity mention; and determining an entity linking result of the at least one entity mention, based on the each embedding vector, the context semantic information, and each type information.

Abstract: A method, apparatus, device, and storage medium for entity linking is disclosed. The method includes: acquiring a target text; determining at least one entity mention included in the target text; determining a candidate entity corresponding to each of the entity mention based on a preset knowledge base; determining a reference text of each of the candidate entity and determining additional feature information of each of the candidate entity; and determining an entity linking result based on the target text, each of the reference text, and each piece of the additional feature information, wherein determining the entity linking result includes determining a probability of linking each of the candidate entity to the entity mention based on a splicing of a first embedding vector and a second embedding vector of the target text and a splicing of a first embedding vector and a second embedding vector of each respective reference text.

Abstract: The present disclosure provides a method, a device, an equipment and a storage medium for mining a topic concept. The method includes: acquiring a plurality of candidate topic concepts based on a query; performing word segmentation on the plurality of candidate topic concepts and performing part-of-speech tagging on words obtained after performing the word segmentation, to obtain a part-of-speech sequence of each of the plurality of candidate topic concepts; and filtering the plurality of candidate topic concepts based on the part-of-speech sequence, to filter out a topic concept corresponding to a target part-of-speech sequence among the plurality of candidate topic concepts, in which a proportion of accurate topic concepts in the target part-of-speech sequence is lower than or equal to a first preset threshold, or a proportion of inaccurate topic concepts in the target part-of-speech sequence is higher than or equal to a second preset threshold.

Abstract: A method of processing multimedia data, a device, and a medium, which relates to a field of an artificial intelligence technology, in particular to fields of knowledge graph and deep learning. The method of processing the multimedia data includes: recognizing the multimedia data so as to obtain at least one key information of the multimedia data; querying a predetermined knowledge base according to the at least one key information, so as to determine a multimedia name associated with the at least one key information and an association degree between the multimedia name and the at least one key information; and determining, in the multimedia name, a name of the multimedia data based on a similarity between alternative multimedia data for the multimedia name and the multimedia data, in response to the association degree being less than a first threshold value.

Abstract: A method of obtaining an event information, an electronic device, and a storage medium, which relate to a field of artificial intelligence, in particular to fields of knowledge graph and deep learning technologies. A specific implementation solution of the method of obtaining the event information includes: determining, according to a query information in data to be processed, a first key information describing an event; determining, according to multimedia data in the data to be processed, a second key information describing an event, wherein the multimedia data includes data obtained by querying based on the query information; and fusing the first key information and the second key information, so as to obtain an event information of a target event described by the data to be processed.

Abstract: Embodiments of the present disclosure relate to a method and apparatus for generating a neural network. The method includes: acquiring a target neural network, the target neural network corresponding to a preset association relationship, and being configured to use two entity vectors corresponding to two entities in a target knowledge graph as an input, to determine whether an association relationship between the two entities corresponding to the inputted two entity vectors is the preset association relationship, the target neural network comprising a relational tensor predetermined for the preset association relationship; converting the relational tensor in the target neural network into a product of a target number of relationship matrices, and generating a candidate neural network comprising the target number of converted relationship matrices; and generating a resulting neural network using the candidate neural network.

Abstract: The present disclosure provides a method and apparatus for acquiring a pre-trained model, an electronic device and a storage medium, and relates to the fields such as deep learning, natural language processing, knowledge graph and intelligent voice. The method may include: acquiring a pre-training task set composed of M pre-training tasks, M being a positive integer greater than 1, the pre-training tasks including: N question-answering tasks corresponding to different question-answering forms, N being a positive integer greater than 1 and less than or equal to M; and jointly pre-training the pre-trained model according to the M pre-training tasks.

Abstract: Technical solutions for video event recognition relate to the fields of knowledge graphs, deep learning and computer vision. A video event graph is constructed, and each event in the video event graph includes: M argument roles of the event and respective arguments of the argument roles, with M being a positive integer greater than one. For a to-be-recognized video, respective arguments of the M argument roles of a to-be-recognized event corresponding to the video are acquired. According to the arguments acquired, an event is selected from the video event graph as a recognized event corresponding to the video.

Abstract: Disclosed are a method for processing multimodal data using a neural network, a device, and a medium, and relates to the field of artificial intelligence and, in particular to multimodal data processing, video classification, and deep learning. The neural network includes: an input subnetwork configured to receive the multimodal data to output respective first features of a plurality of modalities; a plurality of cross-modal feature subnetworks, each of which is configured to receive respective first features of two corresponding modalities to output a cross-modal feature corresponding to the two modalities; a plurality of cross-modal fusion subnetworks, each of which is configured to receive at least one cross-modal feature corresponding to a corresponding target modality and other modalities to output a second feature of the target modality; and an output subnetwork configured to receive respective second features of the plurality of modalities to output a processing result of the multimodal data.

Abstract: A computer-implemented method is provided. The method includes: obtaining a sample text and a sample image corresponding to the sample text; labeling a true semantic tag for the sample text according to a first preset rule; obtaining a text feature representation of the sample text and a predicted semantic tag output by a text coding sub-model; obtaining an image feature representation of the sample image output by an image coding sub-model; calculating a first loss based on the true semantic tag and the predicted semantic tag; calculating a contrast loss based on the text feature representation of the sample text and the image feature representation of the sample image; adjusting parameters of the text coding sub-model based on the first loss and the contrast loss; and adjusting parameters of the image coding sub-model based on the contrast loss.

Abstract: Embodiments of the present disclosure provide a method, apparatus, device and medium for determining text relevance. The method for determining text relevance may include: identifying, from a predefined knowledge base, a first set of knowledge elements associated with a first text and a second set of knowledge elements associated with a second text. The knowledge base includes a knowledge representation consist of knowledge elements. The method may further include: determining knowledge element relevance between the first set of knowledge elements and the second set of knowledge elements, and determining text relevance between the second text and the first text based at least on the knowledge element relevance.

Abstract: A method for generating a pre-trained language model, includes: obtaining sample files; obtaining typography structure information and text information of the sample files by parsing the sample files; obtaining a plurality of task models of a pre-trained language model; obtaining a trained pre-trained language model by jointly training the pre-trained language model and the plurality of task models according to the typography structure information and the text information; and generating a target pre-trained language model by fine-tuning the trained pre-trained language model according to the typography structure information and the text information.

Abstract: The disclosure provides a method for processing a video, an electronic device, and a computer storage medium. The method includes: determining a plurality of first identifiers related to a first object based on a plurality of frames including the first object in a target video; determining a plurality of attribute values associated with the plurality of first identifiers based on a knowledge base related to the first object; determining a set of frames from the plurality of frames, in which one or more attribute values associated with one or more first identifiers determined from each one of the set of frames are predetermined values; and splitting the target video into a plurality of video clips based on positions of the set of frames in the plurality of frames.