iTBLS: A Dataset of Interactive Conversations Over Tabular Information

Anirudh Sundar,Christopher Richardson,William Gay,Larry Heck

2024-04-19

Abstract:This paper introduces Interactive Tables (iTBLS), a dataset of interactive conversations situated in tables from scientific articles. This dataset is designed to facilitate human-AI collaborative problem-solving through AI-powered multi-task tabular capabilities. In contrast to prior work that models interactions as factoid QA or procedure synthesis, iTBLS broadens the scope of interactions to include mathematical reasoning, natural language manipulation, and expansion of existing tables from natural language conversation by delineating interactions into one of three tasks: interpretation, modification, or generation. Additionally, the paper presents a suite of baseline approaches to iTBLS, utilizing zero-shot prompting and parameter-efficient fine-tuning for different computing situations. We also introduce a novel multi-step approach and show how it can be leveraged in conjunction with parameter-efficient fine-tuning to achieve the state-of-the-art on iTBLS; outperforming standard parameter-efficient fine-tuning by up to 15% on interpretation, 18% on modification, and 38% on generation.

Computation and Language,Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

The paper aims to address several key issues related to interaction with tabular information and proposes a dataset named Interactive Tables (iTBLS). Specifically, the goals of the paper include:

1. **Unified Multi-task Capability**: Create a multi-turn dialogue dataset that includes table interpretation, modification, and generation to facilitate collaborative problem-solving between humans and AI. These tasks include understanding table content, manipulating tables through natural language commands, and extending existing tables based on new natural language evidence.

2. **Broaden Interaction Scope**: In addition to previous factual Q&A or process synthesis, iTBLS introduces more complex forms of interaction such as mathematical reasoning, natural language processing, and natural language-based table extensions.

3. **Provide Benchmark Methods**: To handle the various tasks in iTBLS, the paper offers zero-shot prompting, parameter-efficient fine-tuning, and a new multi-step approach to adapt to different computational environments. Specifically, the multi-step approach combined with parameter-efficient fine-tuning improves performance by 15%, 18%, and 38% on interpretation, modification, and generation tasks, respectively.

Through these efforts, the paper seeks to advance research in table data management, particularly in the handling of tabular information in scientific literature.