| With the development of the multimedia and network technology, here comes the era of information explosion. And nowadays the information we obtain is all from where we perceive the world. Compared to text and images, the video can provide more comprehensive and intuitive information, so the video has become the majority media by what users get access to information they need. However, it is very difficult and impractical for the users to find the video what they are interested in from the vast video ocean. The Users need an initiative recommendation system which can help to filter out and present the videos they interested automatically and friendly.This essay studies and implements a novel video recommendation system which is based on knowledge framework and different from the traditional video browsing style to better manage the videos and improve the recommendation accuracy. The contents included in this research are listed as follows:Firstly, we extract the knowledge implicated in the structure behind the video, scene relations, semantic networks and other deep-level contents as video mark by using information extraction technology, and link the target videos together with directed graph by video mark instead of showing them in a simple linear way like it is done in conventional video browsing system. Secondly, we do researches on the similarity measure algorithms and existing recommendation algorithms to find appropriate ones to apply and focus on improving collaborative filtering recommendation to optimize the recommendation accuracy. This video mode has stronger visualization, and it can provide the users with more valuable information in the same unit of time. Finally, we conduct tests and assessment on recommendation accuracy. In the end, the essay gives a summary of research work. It points out the design shortcomings and inadequacies during the research and implement. Meanwhile, it also points out the direction and purpose of future research work. |
PDF Full Text Request