A novel algorithm for extracting text labels and subfigure captions from multi-panel figure caption

Segmentation of multi-panel figure caption into subfigure captions and text labels is a problem that arises in a number of applications particularly in estimating the total number of panels in the multi-panel figure. The results of the existing methods for solving this problem are unsatisfactory and need improvement. Moreover, these methods are computationally expensive. In this paper, we propose a novel algorithm for the segmentation of multi-panel figure caption. This algorithm needs two inputs, i.e., multi-panel figure caption to be segmented and text label type. Upon receiving these two inputs it will extracts text labels and subfigure captions from the given multi-panel figure caption. We implemented the proposed algorithm and evaluated its results on 100 multipanel figure captions collected from 100 research papers. From the results we observed 100% accuracy for the text label types containing two or more than two characters.