PlotQA: Reasoning over Scientific Plots

TL;DR

PlotQA tackles the gap in plot reasoning by introducing a real-world, large-scale QA dataset with open vocabulary answers and numeric reasoning, grounded on 224,377 plots and 28.95 million questions generated from 74 templates. It also proposes a hybrid model that splits questions into fixed-vocabulary and open-vocabulary paths, the latter flowing through a four-stage perception-to-QA pipeline (VED, OCR, SIE, Table QA) to reason over a constructed knowledge graph. The results show that existing VQA models struggle with open-ended numeric questions, while the proposed approach improves DVQA performance and achieves the best reported PlotQA accuracy to date, albeit still below human performance and highlighting the need for more accurate visual element detection and robust reasoning components. Overall, PlotQA advances the field by providing a more realistic benchmark and a modular reasoning framework that can be extended toward broader, real-world plot understanding.

Abstract

Existing synthetic datasets (FigureQA, DVQA) for reasoning over plots do not contain variability in data labels, real-valued data, or complex reasoning questions. Consequently, proposed models for these datasets do not fully address the challenge of reasoning over plots. In particular, they assume that the answer comes either from a small fixed size vocabulary or from a bounding box within the image. However, in practice, this is an unrealistic assumption because many questions require reasoning and thus have real-valued answers which appear neither in a small fixed size vocabulary nor in the image. In this work, we aim to bridge this gap between existing datasets and real-world plots. Specifically, we propose PlotQA with 28.9 million question-answer pairs over 224,377 plots on data from real-world sources and questions based on crowd-sourced question templates. Further, 80.76% of the out-of-vocabulary (OOV) questions in PlotQA have answers that are not in a fixed vocabulary. Analysis of existing models on PlotQA reveals that they cannot deal with OOV questions: their overall accuracy on our dataset is in single digits. This is not surprising given that these models were not designed for such questions. As a step towards a more holistic model which can address fixed vocabulary as well as OOV questions, we propose a hybrid approach: Specific questions are answered by choosing the answer from a fixed vocabulary or by extracting it from a predicted bounding box in the plot, while other questions are answered with a table question-answering engine which is fed with a structured table generated by detecting visual elements from the image. On the existing DVQA dataset, our model has an accuracy of 58%, significantly improving on the highest reported accuracy of 46%. On PlotQA, our model has an accuracy of 22.52%, which is significantly better than state of the art models.

Figures (12)

• Figure 1: A sample {plot, question, answer} triplet from FigureQA, DVQA, and PlotQA (our) datasets.
• Figure 2: Sample plots of different types in the PlotQA dataset.
• Figure 3: Our proposed multi-staged modular pipeline for QA on scientific plots.
• Figure 4: Our proposed model containing (i) a question classifier for deciding whether the question can be answered from a fixed vocabulary (orange) or needs more complex reasoning (green), (ii) QA-as-classification model to answer questions of the former type, and (iii) multi-staged model as a pipeline of perception and QA modules for answering complex questions.
• Figure 5: Ground-truth (cyan) and predicted (red) boxes.