Leveraging Query Resolution and Reading Comprehension for Conversational Passage Retrieval

This paper describes the participation of UvA.ILPS group at the TREC CAsT\n2020 track. Our passage retrieval pipeline consists of (i) an initial retrieval\nmodule that uses BM25, and (ii) a re-ranking module that combines the score of\na BERT ranking model with the score of a machine comprehension model adjusted\nfor passage retrieval. An important challenge in conversational passage\nretrieval is that queries are often under-specified. Thus, we perform query\nresolution, that is, add missing context from the conversation history to the\ncurrent turn query using QuReTeC, a term classification query resolution model.\nWe show that our best automatic and manual runs outperform the corresponding\nmedian runs by a large margin.\n