OpenCQA: Open-ended Question Answering with Charts

29854
Back to Articles Dashboard

Authors

nan

Abstract

AbstractCharts are very popular to analyze data and convey important insights. People often analyze visualizations to answer open-ended questions that require explanatory answers. Answering such questions are often difficult and time-consuming as it requires a lot of cognitive and perceptual efforts. To address this challenge, we introduce a new task called OpenCQA, where the goal is to answer an open-ended question about a chart with descriptive texts. We present the annotation process and an in-depth analysis of our dataset. We implement and evaluate a set of baselines under three practical settings. In the first setting, a chart and the accompanying article is provided as input to the model. The second setting provides only the relevant paragraph(s) to the chart instead of the entire article, whereas the third setting requires the model to generate an answer solely based on the chart. Our analysis of the results show that the top performing models generally produce fluent and coherent text while they struggle to perform complex logical and arithmetic reasoning.

Keywords

nan

Full Content

Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 11817 - 11837 December 7-11, 2022 ©2022 Association for Computational Linguistics                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             …

Content truncated to 500 words

Labeled Text

AbstractCharts are very popular to analyze data and convey important insights. People often analyze visualizations to answer open-ended questions that require explanatory answers. Answering such questions are often difficult and time-consuming as it requires a lot of cognitive and perceptual efforts. To address this challenge, we introduce a new task called OpenCQA, where the goal is to answer an open-ended question about a chart with descriptive texts. We present the annotation process and an in-depth analysis of our dataset. We implement and evaluate a set of baselines under three practical settings. In the first setting, a chart and the accompanying article is provided as input to the model. The second setting provides only the relevant paragraph(s) to the chart instead of the entire article, whereas the third setting requires the model to generate an answer solely based on the chart. Our analysis of the results show that the top performing models generally produce fluent and coherent text while they struggle to perform complex logical and arithmetic reasoning.

Publication Details

Paper ID:
29854
Year:
2022
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan

Links

Paper URL PDF Publication URL