TAPAS: Weakly Supervised Table Parsing via Pre-training

Abstract

Answering natural language questions over tables is usually seen as a semantic parsing task. To alleviate the collection cost of full logical forms, one popular approach focuses on weak supervision consisting of denotations instead of logical forms. However, training semantic parsers from weak supervision poses difficulties, and in addition, the generated logical forms are only used as an intermediate step prior to retrieving the denotation. In this paper, we present TAPAS, an approach to question answering over tables without generating logical forms. TAPAS trains from weak supervision, and predicts the denotation by selecting table cells and optionally applying a corresponding aggregation operator to such selection. TAPAS extends BERT's architecture to encode tables as input, initializes from an effective joint pre-training of text segments and tables crawled from Wikipedia, and is trained end-to-end. We experiment with three different semantic parsing datasets, and find that TAPAS outperforms or rivals semantic parsing models by improving state-of-the-art accuracy on SQA from 55.1 to 67.2 and performing on par with the state-of-the-art on WIKISQL and WIKITQ, but with a simpler model architecture. We additionally find that transfer learning, which is trivial in our setting, from WIKISQL to WIKITQ, yields 48.7 accuracy, 4.2 points above the state-of-the-art.

Accepted to ACL 2020

References in corpus (9)

• Cross-lingual Language Model Pretraining
• Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning
• Semi-supervised Sequence Learning
• TabFact: A Large-scale Dataset for Table-based Fact Verification
• A Comprehensive Exploration on WikiSQL with Table-Aware Word Contextualization
• Learning to Generalize from Sparse and Underspecified Rewards
• Learning a Natural Language Interface with Neural Programmer
• Adversarial TableQA: Attention Supervision for Question Answering on Tables
• Giving BERT a Calculator: Finding Operations and Arguments with Reading Comprehension

Cited by in corpus (24)

• NL4DV: A Toolkit for Generating Analytic Specifications for Data Visualization from Natural Language Queries
• Towards Natural Language Interfaces for Data Visualization: A Survey
• TAPEX: Table Pre-training via Learning a Neural SQL Executor
• Annotating Columns with Pre-trained Language Models
• GraPPa: Grammar-Augmented Pre-Training for Table Semantic Parsing
• TURL: Table Understanding through Representation Learning
• Towards Complex Document Understanding By Discrete Reasoning
• Document AI: Benchmarks, Models and Applications
• Retrieving Complex Tables with Multi-Granular Graph Representation Learning
• On Grounded Planning for Embodied Tasks with Language Models
• Language Models as Knowledge Bases: On Entity Representations, Storage Capacity, and Paraphrased Queries
• Stage-wise Fine-tuning for Graph-to-Text Generation
• A Unified Transferable Model for ML-Enhanced DBMS
• KGPT: Knowledge-Grounded Pre-Training for Data-to-Text Generation
• Around the GLOBE: Numerical Aggregation Question-Answering on Heterogeneous Genealogical Knowledge Graphs with Deep Neural Networks
• Meta-Learning for Domain Generalization in Semantic Parsing
• Data Augmentation with Hierarchical SQL-to-Question Generation for Cross-domain Text-to-SQL Parsing
• Open Question Answering over Tables and Text
• A Graph Representation of Semi-structured Data for Web Question Answering
• WikiTableT: A Large-Scale Data-to-Text Dataset for Generating Wikipedia Article Sections
• Sattiy at SemEval-2021 Task 9: An Ensemble Solution for Statement Verification and Evidence Finding with Tables
• Dual Reader-Parser on Hybrid Textual and Tabular Evidence for Open Domain Question Answering
• A question-answering system for aircraft pilots' documentation
• Data-QuestEval: A Referenceless Metric for Data-to-Text Semantic Evaluation