pdf bibtex 1 citation workshop paper
Appeared in: The First Workshop on Knowledge-Enhanced Information Retrieval (KEIR@ECIR 2024)
Abstract:
Despite considerable progress in neural relevance ranking techniques, search engines still struggle to process complex queries effec- tively — both in terms of precision and recall. Sparse and dense Pseudo-Relevance Feedback (PRF) approaches have the potential to overcome limitations in recall, but are only effective with high precision in the top ranks. In this work, we tackle the problem of search over complex queries using three complementary techniques. First, we demonstrate that applying a strong neural re-ranker before sparse or dense PRF can improve the retrieval effectiveness by 5–8%. Second, we propose an enhanced expan- sion model, Latent Entity Expansion (LEE), which applies fine-grained word and entity-based relevance modelling incorporating localized fea- tures. Specifically, we find that by including both words and entities for expansion achieve a further 2–8% improvement in NDCG. Our analysis also demonstrates that LEE is largely robust to its parameters across datasets and performs well on entity-centric queries. And third, we in- clude an “adaptive” component in the retrieval process, which iteratively refines the re-ranking pool during scoring using the expansion model and avoids re-ranking additional documents. We find that this combination of techniques achieves the best NDCG, MAP and R@1k results on the TREC Robust 2004 and CODEC document datasets.
BibTeX @inproceedings{mackie:keir2024-lee, author = {Mackie, Iain and Chatterjee, Shubham and MacAvaney, Sean and Dalton, Jeff}, title = {Adaptive Latent Entity Expansion for Document Retrieval}, booktitle = {The First Workshop on Knowledge-Enhanced Information Retrieval}, year = {2024}, url = {https://arxiv.org/abs/2306.17082} }