Natural Language Processing for Recommender Systems

Recommender systems have emerged as pervasive tools helping people to navigate vast amounts of available information, personalize search results and actively suggest novel contents of potential interest to users. A variety of signals and techniques are used to elicit user preferences for the best match with their current and potential future interests. Well-established techniques, such as collaborative filtering exploit user and item similarity, typically from a description of recommended subjects, user reviews, and ranking. Many of these sources are in textual form that contain a tremendous amount of information about subjects and objects of recommendations. While these signals have already been utilized, recent advancements in machine learning, and especially deep learning, allow the discovery of contextual interactions and semantical similarities. Moreover, pretrained language models, such as BERT and GPT-3 improve textual representation, help to resolve ambiguities, and shorten the processing time.

In this Research Topic collection, we welcome publications from researchers in both academia and industry on the latest developments on the intersection between Natural Language Processing and Recommender Systems. The submissions will range from theoretically motivated new algorithms to empirically validated solutions and user studies.

Topics of interest include, but are not limited to, the following:

• Augmenting recommendation techniques with signals derived from user generated textual reviews

• Advances in the research of conversational recommenders

• Natural Language Processing techniques for increased interpretability and explainability of recommender models

• Sentiment and aspect analysis for Recommender Systems

• Language models for Recommender Systems

• Summarisation of product reviews