Improving media content recommendation with automatic annotations

With the immense growth of media content production on the internet and increasing wariness about privacy, content-based recommendation systems offer the possibility of promoting media to users (e.g. posts, videos, podcasts) based solely on  a representation of the content, i.e. without using any user-related data such as views and more generally interactions between users and items. In this work, we study the potential of using off-the-shelf automatic annotation tools from the Information Extraction literature to improve recommendation performance without any extra cost of training, data collection or annotation. We experiment with how these annotations can improve recommendations on two tasks: the traditional user history-based recommendation, as well as a purely content-based recommendation evaluation. We pair these automatic annotations with the manually created metadata and we show that Knowledge Graphs through their embeddings constitute a great modality to seamlessly integrate this extracted knowledge and provide better recommendations. The evaluation code, as well as the enrichment generation, is available at https://github.com/D2KLab/ka-recsys.