Capturing Row and Column Semantics in Transformer Based Question Answering over Tables

Abstract

Transformer based architectures are recently used for the task of answering questions over tables. In order to improve the accuracy on this task, specialized pre-training techniques have been developed and applied on millions of open-domain web tables. In this paper, we propose two novel approaches demonstrating that one can achieve superior performance on table QA task without even using any of these specialized pre-training techniques. The first model, called RCI interaction, leverages a transformer based architecture that independently classifies rows and columns to identify relevant cells. While this model yields extremely high accuracy at finding cell values on recent benchmarks, a second model we propose, called RCI representation, provides a significant efficiency advantage for online QA systems over tables by materializing embeddings for existing tables. Experiments on recent benchmarks prove that the proposed methods can effectively locate cell values on tables (up to ∼98% Hit@1 accuracy on WikiSQL lookup questions). Also, the interaction model outperforms the state-of-the-art transformer based approaches, pre-trained on very large table corpora (TAPAS and TABERT), achieving ∼3.4% and ∼18.86% additional precision improvement on the standard WikiSQL benchmark 1 .

DiSCoMaT: Distantly Supervised Composition Extraction from Tables in Material Science Articles

“…Recent work has developed neural models for various NLP tasks based on tabular data, for instance, tabular natural language inference [40,29], QA over one or a corpus of tables [14,56,1,8,41,3], table orientation classification [12,38], and relation extraction from tables [9,26]. Several papers study QA models.…”

“…TaBERT [56] uses a different pre-training objective and model architecture to jointly learn NL sentence and table representations. Row and column intersection (RCI) model [8] predicts the probability of finding the answer to a given question in each row and column of a table independently, without any pre-training on tables. TaPEx [25] uses a BART encoder-decoder architecture, pre-trained via synthetic executable programs to mimic the behaviour of SQL execution engine.…”