A Question-Answering framework for plots using Deep learning

  • 2018-06-12 17:31:23
  • Revanth Reddy, Rahul Ramesh, Ameet Deshpande, Mitesh M. Khapra
  • 0

Abstract

Deep Learning has managed to push boundaries in a wide variety of tasks. Onearea of interest is to tackle problems in reasoning and understanding, in anaim to emulate human intelligence. In this work, we describe a deep learningmodel that addresses the reasoning task of question-answering on bar graphs andpie charts. We introduce a novel architecture that learns to identify variousplot elements, quantify the represented values and determine a relativeordering of these statistical values. We test our model on the recentlyreleased FigureQA dataset, which provides images and accompanying questions,for bar graphs and pie charts, augmented with rich annotations. Our approachoutperforms the state-of-the-art Relation Networks baseline and traditionalCNN-LSTM models when evaluated on this dataset. Our model also has aconsiderably faster training time of approximately 2 days on 1 GPU compared tothe Relation Networks baseline

 

Introduction (beta)

None

 

Conclusion (beta)

None