Lightweight Tag-Aware Personalized Recommendation on the Social Web Using Ontological Similarity

With the rapid growth of social tagging systems, many research efforts are being put into personalized search and recommendation using social tags (i.e., folksonomies). As users can freely choose their own vocabulary, social tags can be very ambiguous (for instance, due to the use of homonyms or synonyms). Machine learning techniques (such as clustering and deep neural networks) are usually applied to overcome this tag ambiguity problem. However, the machine-learning-based solutions always need very powerful computing facilities to train recommendation models from a large amount of data, so they are inappropriate to be used in lightweight recommender systems. In this paper, we propose an ontological similarity to tackle the tag ambiguity problem without the need of model training by using contextual information. The novelty of this ontological similarity is that it first leverages external domain ontologies to disambiguate tag information, and then semantically quantifies the relevance between user and item profiles according to the semantic similarity of the matching concepts of tags in the respective profiles. Our experiments show that the proposed ontological similarity is semantically more accurate than the state-of-the-art similarity metrics, and can thus be applied to improve the performance of content-based tag-aware personalized recommendation on the social web. Consequently, as a model-training-free solution, ontological similarity is a good disambiguation choice for lightweight recommender systems and a complement to machine-learning-based recommendation solutions.