Neural Endorsement Based Contextual Suggestion

This paper presents the University of Amsterdam’s participation in the TREC 2016 Contextual Suggestion Track. In this research, we have studied a personallized neural document language modeling and a neural category preference modeling for contextual suggestion using available endorsements in TREC 2016 contextual suggestion track phase 2 requests. Specifically, our main aim is to answer the questions: How to model users’ profiles by using the suggestions’ endorsements as an additional data? How effective is using word embeddings to boost terms’ weights relevant to the given endorsements? How to model users’ attractioncategory preferences? How effective is using deep neural networks to learn users’ category preferences in contextual suggestion task? Our main findings are the following: First, the neural personalized document based user profiling using word embeddings improves the baseline content-based filtering approach based on all the common IR measures including TREC 2016 Contextual Suggestion official metric (NDCG@5). Second, neural users’ category preference modeling beats both baseline content-based filtering and the user profiling model using word-embeddings in terms of all the common IR measures.