Duet at TREC 2019 Deep Learning Track
This work addresses information retrieval challenges for researchers and practitioners, but it is incremental as it builds on existing Duet methods with minor adaptations.
The authors tackled document and passage retrieval tasks at TREC 2019 by adapting the Duet architecture, introducing DuetMF for multiple field document views and using ensembles and learning-to-rank, achieving improved performance over baselines.
This report discusses three submissions based on the Duet architecture to the Deep Learning track at TREC 2019. For the document retrieval task, we adapt the Duet model to ingest a "multiple field" view of documents---we refer to the new architecture as Duet with Multiple Fields (DuetMF). A second submission combines the DuetMF model with other neural and traditional relevance estimators in a learning-to-rank framework and achieves improved performance over the DuetMF baseline. For the passage retrieval task, we submit a single run based on an ensemble of eight Duet models.