论文信息 - TUA1 at the TREC 2019: Deep Learning Track

TUA1 at the TREC 2019: Deep Learning Track

This paper details our participation in the TREC 2019 Deep Learning Track task including Passage Ranking task. In the Top-1000 Re-ranking subtask of Passage Ranking task, we performed passage ranking based on the technique of Conv-KNRM and BERT. In order to get a better ranking result, we divided the task into two stages. In the first stage we use Conv-KNRM model for initial ranking, then in the second stage we combine the results with a state-of-the-art BERT re-ranker.

[1] Stephen E. Robertson,et al. GatfordCentre for Interactive Systems ResearchDepartment of Information , 1996 .

[2] Jimmy J. Lin,et al. Document Expansion by Query Prediction , 2019, ArXiv.

[3] Zhiyuan Liu,et al. Convolutional Neural Networks for Soft-Matching N-Grams in Ad-hoc Search , 2018, WSDM.

[4] Jianfeng Gao,et al. A Human Generated MAchine Reading COmprehension Dataset , 2018 .

[5] Ming-Wei Chang,et al. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding , 2019, NAACL.

[6] Bhaskar Mitra,et al. Overview of the TREC 2019 deep learning track , 2020, ArXiv.