Knowledge Graph: A Recommendation Method for New Media Short Videos

Abstract

The recommendation of new media short videos plays a crucial role in today's social media and content consumption. Short videos can rapidly disseminate information, capture users' attention, and quickly accumulate a large audience. According to the latest data, as of April 2024, TikTok, a globally renowned short video social media platform, has been downloaded over 4.92 billion times worldwide, with more than 1.582 billion monthly active users. However, the recommendation algorithms of current mainstream social platforms often favor content that generates a large amount of interaction in the short term, which may negatively impact the dissemination of content with long-term value and depth. To address this challenge, this paper proposes a knowledge graph-based recommendation method for new media short videos, which structures the representation of short video content by associating it with related concepts, entities (such as people, places, events, etc.), and their relationships, thereby providing a more accurate understanding of the video content’s semantics. This paper delves into key technologies including knowledge graph construction, user interest modeling, short video content modeling, semantic matching, and recommendation. Through practical application cases and comparative experiments, the effectiveness of this method is evaluated. The results show that the application of knowledge graphs in short video recommendation can significantly improve the quality of content recommendations, offering users a richer and more personalized viewing experience.