Overview of the TREC 2020 deep learning track

TL;DR

This paper summarizes the second year of the TREC Deep Learning Track, focusing on ad hoc ranking with large training datasets, and confirms that BERT-style pretraining yields superior performance in this regime.

Contribution

It provides an overview of the TREC 2020 deep learning track, highlighting the effectiveness of BERT-style models in large-scale ranking tasks.

Findings

BERT-style rankers outperform others in large data settings

Evaluation with TREC-style metrics provides insights into ranking method effectiveness

Large annotated datasets improve ranking model performance

Abstract

This is the second year of the TREC Deep Learning Track, with the goal of studying ad hoc ranking in the large training data regime. We again have a document retrieval task and a passage retrieval task, each with hundreds of thousands of human-labeled training queries. We evaluate using single-shot TREC-style evaluation, to give us a picture of which ranking methods work best when large data is available, with much more comprehensive relevance labeling on the small number of test queries. This year we have further evidence that rankers with BERT-style pretraining outperform other rankers in the large data regime.