Open Domain Question Answering over Tables via Dense Retrieval

2021/03/22 by Herzig, Jonathan, Müller, Thomas, Krichene, Syrine +1 · 8 citations
#Computation and Language (cs.CL) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2103.12011

Abstract

Recent advances in open-domain QA have led to strong models based on dense retrieval, but only focused on retrieving textual passages. In this work, we tackle open-domain QA over tables for the first time, and show that retrieval can be improved by a retriever designed to handle tabular context. We present an effective pre-training procedure for our retriever and improve retrieval quality with mined hard negatives. As relevant datasets are missing, we extract a subset of Natural Questions (Kwiatkowski et al., 2019) into a Table QA dataset. We find that our retriever improves retrieval results from 72.0 to 81.1 recall@10 and end-to-end QA results from 33.8 to 37.7 exact match, over a BERT based retriever.

Cited by

• TabRank: Chain-of-Thought Distillation for Table Re-Rankers
• Exploring Multi-Table Retrieval Through Iterative Search
• HyFedRAG: A Federated Retrieval-Augmented Generation Framework for Heterogeneous and Privacy-Sensitive Data
• Granite Embedding R2 Models
• Improving Table Retrieval with Question Generation from Partial Tables
• Empowering Tabular Data Preparation with Language Models: Why and How?
• RASL: Retrieval Augmented Schema Linking for Massive Database Text-to-SQL
• Rethinking Dataset Discovery with DataScout

Related

• Enhancing Open-Domain Table Question Answering via Syntax- and Structure-aware Dense Retrieval
• CRAFT: Training-Free Cascaded Retrieval for Tabular QA
• Denoising Table-Text Retrieval for Open-Domain Question Answering
• Improving Table Retrieval with Question Generation from Partial Tables
• Dense Passage Retrieval for Open-Domain Question Answering