Method for retrieving specific action video fragments in Japanese online video corpus

Abstract

The invention discloses a method for retrieving a specific action video fragment in a Japanese online video corpus, belongs to the field of video fragment retrieval, and relates to a method for quickly retrieving a video fragment containing a specific action in an online Japanese multi-modal corpus based on a deep learning technology. The retrieval method adopts a deep learning and statistical learning technology, an action and subtitle and video matching positioning technology, a caching and indexing technology and a data visualization technology. The retrieval method comprises the four steps of uploading of video corpora, frame-by-frame action analysis and feature extraction of the video corpora, index establishment and action retrieval. Firstly, after uploading of video corpora is finished, an action analysis and data processing server carries out analysis and feature extraction on the video corpora, a final result is indexed, and a user carries out retrieval through a WEB application. According to the method, quick query, accurate positioning and downloading of video fragment retrieval of specific actions in Japanese video corpora are realized, and a convenient retrieval service is provided for Japanese learning and research.

Description

technical field [0001] The invention belongs to the field of video clip retrieval, and relates to a fast retrieval method for video clips containing specific actions in an online Japanese multimodal corpus based on deep learning technology. Background technique [0002] In recent years, with the development of Internet technology, more and more foreign language learners try to learn foreign languages ​​through video corpus. Video corpus has received high attention in the process of foreign language learning, because video corpus can make up for the lack of pure text corpus. Provide a more realistic context. Actions play an important role in video corpus, because actions are often the embodiment of context and culture. Finding actions that appear in videos is of great significance to both foreign language learners and researchers. For Japanese learners, it is helpful to understand the specific context and culture of a language, and to deepen the impression of vocabulary, gra...

AI Technical Summary

Problems solved by technology

[0003] At present, although there are Japanese video corpora, such as the patent of Jiang Guohai et al., ZL201310216448.2 "Video Segment Retrieval Method for Japanese Online Video Corpus", none of them have the functions of action retrieval and feature extraction, and the rich features brought by actions are lost. Semantic information makes it difficult to carry out tasks such as context classification, pragmatic research, and cultural analysis

Images