[This Bachelor/Master thesis can be written either in German or English.]

Nowadays, recommender systems (RS) are a pervasive presence in everyday life, from streaming services to any kind of social apps. To date, and aiming to decrease the perception of RS as black boxes by users, various explanatory methods and models have been developed. These methods are limited - to a greater extent - to provide static explanations, while interactive explanations (those which allow users to question the system for additional reasons of why an item is recommended) seem to be a promising field to explore. Particularly, we believe that - under certain circumstances - dialog-based explanations could be better perceived by users, compared to static explanations, in terms of transparency and effectiveness. On the other hand, conversational interfaces (e.g. chatbots) have gained popularity in recent years, allowing users to interact in a flexible and natural manner with a system. We therefore aim to explore the possibilities and advantages of using conversational agents as part of an explanatory RS, particularly systems that leverage user reviews to generate both recommendations and explanations, given the richness of their content.

In consequence, we aim with this proposed thesis to explore and design a conversational explanatory interface that would enable an explanatory dialogue between a user and the review-based RS. A second aim of this project would be the development and testing of a small prototype to validate such design.

Programming skills (Python preferred) are desirable, knowledge in text mining, machine learning or deep learning is a great plus.

If you are interested in this work, please contact Diana C. Hernandez Bocanegra (enclosing a current overview of grades).