Scientific Highlight “Modelling Cognitive Processes for Tag Recommendations”

Social Computing

The novel cognitive-inspired tag recommendation approach based on activation processes in human memory provides better recommendation accuracy results than state-of-the-art algorithms.

Our first attempt in the research direction of theory-inspired recommender systems has been the development of a tag recommendation approach based on activation processes in human memory. Therefore, in (Kowald, 2015 KC; Kowald & Lex, 2016 KC), we investigated the relationship between activation processes in human memory and the reuse of tags in six social tagging systems. Here, we found that (i) past usage frequency, (ii) recency, and (iii) the current semantic context are important factors when people reuse tags, which corresponds to the theory of activation processes in human memory. Next, in (Kowald & Lex, 2015 KC), we built upon these results in order to utilize the activation equation of the cognitive architecture ACT-R for designing, evaluating and implementing a novel tag recommendation algorithm. Finally, in (Kowald, Pujari & Lex, 2017 KC), we generalized this approach for related use cases in the area of tag-based recommender systems such as hashtag recommendations in Twitter. Our cognitive-inspired approach termed BLLI,S,C provided better recommendation accuracy results than current state-of-the-art algorithms, which are designed in a more data-driven way.