Key Research Highlights > Databases & Big Data Analytics > Personalised Search and Recommendations Using Tags
Mong Li LEE
Personalised Search and Recommendations Using Tags
Users of the social graph use tags to annotate and share content with their friends. A standing challenge is how best to incorporate this user generated information to retrieve relevant resources for querying users.
To address the objective of this project, we have developed an incremental algorithm that builds a social network graph identifying authoritative users for queries submitted. Authoritative users are those who have previously tagged content that is relevant to the issued query. Additionally, we have proposed a model based on 4-order tensor and higher order singular value decomposition to reveal latent semantic associations among users, items, tags and rating to overcome limitations of ternary systems typical of recommender systems. Finally, we developed a system to provide online relevance feedback to increase user acceptance of recommendations using tags.