Abstract: The present disclosure provides a method and apparatus for constructing an event library, which relate to the technical fields of deep learning, natural language processing, big data, and the like. An implementation includes: acquiring at least one event text data, the at least one event text data being to be assigned to a text library; obtaining an extraction event name based on the event text data; performing a match between the extraction event name and event information in the text library, to obtain a recalled event in the text library; detecting, based on the recalled event, whether the extraction event name meets a unifying condition; and in response to detecting that the extraction event name does not meet the unifying condition, obtaining a new event name in the event library based on the extraction event name, and adding the event text data as a new event into the text library.

Abstract: The present disclosure discloses a method and apparatus for expanding a query. The method comprises: calculating an image click characteristic vector of a target query based on an image click log associated with the target query; finding a similar query of the target query based on the image click characteristic vector, to obtain a candidate expansion query set of the target query; matching the target query and each candidate expansion query in the candidate expansion query set with a knowledge graph, to extract an entity word and a qualifier of the target query and an entity word and a qualifier of the each candidate expansion query; expanding the entity words and the qualifiers of the target query and the each candidate expansion query in combination with the knowledge graph; and matching using expansion results of the entity words and the qualifiers, to determine an expanded query of the target query.

Abstract: A method for incrementally training a model is provided. The method includes: obtaining a first soft tag of each raw sample in a raw sample set based on a basic target model trained using the raw sample set; selecting a plurality of raw samples from the raw sample set as first samples based on the first soft tag of each raw sample; obtaining second samples and a second soft tag of each second sample; determining the first samples, the respective first soft tags, the second samples and the respective second soft tags as training samples; and generating a final target model by inputting the training samples into the basic target model for model training.

Abstract: A method, an apparatus, a device and a storage medium for embedding user app interest are provided. The method includes: acquiring a user existing app installation list and a user app installation list within a predetermined time window, where the app includes app ID information and app category information; inputting the existing app installation list and the app installation list within the predetermined time window into a pre-trained user app interest embedding model to obtain a user app interest embedding vector. By combining the user existing app installation list information and the user recent app installation list information, the user app interest embedding vector may simultaneously reflect the user long-term interest and the user short-term interest.

Abstract: A method for retrieving an image-text block from a web page is provided, which comprises: retrieving an image node; filtering the image node to obtain candidate image nodes; traversing, for each of the candidate image nodes, a node in sequence toward an ancestor node of the candidate image node in a preset maximum traversal depth until an ancestor node with a text is visited, using the ancestor node with the text as a candidate image-text block; clustering the candidate image-text blocks based on hash values of the path information of the candidate image-text blocks; and determining, for each image-text block cluster, a common ancestor node of the candidate image-text blocks within the image-text block cluster based on the path information of the candidate image-text blocks, and determining path information of the image-text block cluster based on the common ancestor node.

Abstract: A method for retrieving an image-text block from a web page is provided, which comprises: retrieving an image node; filtering the image node to obtain candidate image nodes; traversing, for each of the candidate image nodes, a node in sequence toward an ancestor node of the candidate image node in a preset maximum traversal depth until an ancestor node with a text is visited, using the ancestor node with the text as a candidate image-text block; clustering the candidate image-text blocks based on hash values of the path information of the candidate image-text blocks; and determining, for each image-text block cluster, a common ancestor node of the candidate image-text blocks within the image-text block cluster based on the path information of the candidate image-text blocks, and determining path information of the image-text block cluster based on the common ancestor node.

Abstract: The present disclosure discloses a method and apparatus for expanding a query. The method comprises: calculating an image click characteristic vector of a target query based on an image click log associated with the target query; finding a similar query of the target query based on the image click characteristic vector, to obtain a candidate expansion query set of the target query; matching the target query and each candidate expansion query in the candidate expansion query set with a knowledge graph, to extract an entity word and a qualifier of the target query and an entity word and a qualifier of the each candidate expansion query; expanding the entity words and the qualifiers of the target query and the each candidate expansion query in combination with the knowledge graph; and matching using expansion results of the entity words and the qualifiers, to determine an expanded query of the target query.