A Line in the Sand: Recommendation or Ad-hoc Retrieval?
This work addresses music recommendation for platforms like Spotify by proposing an incremental approach that adapts existing retrieval techniques to a new domain.
The paper tackled the problem of music recommendation by framing it as an ad-hoc retrieval task, combining standard retrieval models with pseudo-relevance feedback and content-based methods, achieving competitive results such as rank 7 out of 112 teams in the RecSys 2018 challenge.
The popular approaches to recommendation and ad-hoc retrieval tasks are largely distinct in the literature. In this work, we argue that many recommendation problems can also be cast as ad-hoc retrieval tasks. To demonstrate this, we build a solution for the RecSys 2018 Spotify challenge by combining standard ad-hoc retrieval models and using popular retrieval tools sets. We draw a parallel between the playlist continuation task and the task of finding good expansion terms for queries in ad-hoc retrieval, and show that standard pseudo-relevance feedback can be effective as a collaborative filtering approach. We also use ad-hoc retrieval for content-based recommendation by treating the input playlist title as a query and associating all candidate tracks with meta-descriptions extracted from the background data. The recommendations from these two approaches are further supplemented by a nearest neighbor search based on track embeddings learned by a popular neural model. Our final ranked list of recommendations is produced by a learning to rank model. Our proposed solution using ad-hoc retrieval models achieved a competitive performance on the music recommendation task at RecSys 2018 challenge---finishing at rank 7 out of 112 participating teams and at rank 5 out of 31 teams for the main and the creative tracks, respectively.