Simple Applications of BERT for Ad Hoc Document Retrieval
This work addresses document retrieval for search applications, but it is incremental as it adapts an existing method to a specific bottleneck.
The paper tackled the challenge of applying BERT to ad hoc document retrieval by handling long documents through sentence-level inference and score aggregation, achieving the highest average precision on TREC microblog and newswire datasets among known neural approaches.
Following recent successes in applying BERT to question answering, we explore simple applications to ad hoc document retrieval. This required confronting the challenge posed by documents that are typically longer than the length of input BERT was designed to handle. We address this issue by applying inference on sentences individually, and then aggregating sentence scores to produce document scores. Experiments on TREC microblog and newswire test collections show that our approach is simple yet effective, as we report the highest average precision on these datasets by neural approaches that we are aware of.