Extraction and interpretation of charts in technical documents

Abstract

The Information Extraction is a method for filtering information from large volumes of text. It includes the extraction of documents from collections and the tagging of particular terms in text. But non-text information such as graphs, images, figures, etc are common in any technical documents. Scientific charts are commonly used in graphical representation of statistical, experimental and technical data. These are the major visual aids for data analysis and are simple, clear and widely used. Image understanding is the research area concerned with the design and experimentation of computer systems that integrate models of a visual image problem domain. A system for recognition and interpretation of simple bar graphs is proposed. The system generates a natural language description based on semantic understanding. An approach for chart interpretation and conversion to natural language text is discussed. Chart interpretation processes graphical and textual components separately, and then it associates the graphical information to its corresponding textual data. The semantic meaning of the chart is presented in a natural language format.

DVQA: Understanding Data Visualizations via Question Answering

“…Extracting data from bar charts using computer vision has been extensively studied [2,6,19,29,31]. Some focus on extracting the visual elements from the bar charts [29], while others focus on extracting the data from each bar directly [31,19].…”

“…Extracting data from bar charts using computer vision has been extensively studied [2,6,19,29,31]. Some focus on extracting the visual elements from the bar charts [29], while others focus on extracting the data from each bar directly [31,19]. Most of these approaches use fixed heuristics and make strong simplifying assumptions, e.g., [31] made several simplifying assumptions about bar chart appearance (bars are solidly shaded without textures or gradients, no stacked bars, etc.).…”

Accessible charts are part of the equation of accessible papers: a heuristic evaluation of the highest impact LIS journals

“…Similarly, but oriented to a broader set of image types, the work of Splendiani (2015) focuses on how to textually describe non-text content for scientific articles. On the other hand, authors such as Corio and Lapalme (1999), Chester and Elzer (2005), Elzer et al (2008), Ferres et al (2010), Greenbackeret al (2011, Gao et al (2012a), Nazemi and Murray (2013), Kallimani et al (2013) or De (2018) propose different methods for the automated generation of textual alternatives from the information available in a chart. For their part, authors such as Elzer et al (2007), Agarwal and Yu (2009) or Yu et al (2009) have studied the importance of captions for the understanding of a chart as "it often concisely summarizes a paper's most important results" (Cohen et al, 2003).…”

An Approach for Chart Description Generation in Cyber–Physical–Social System

“…Mahmood and Bajwa [29] used a template-based NLG approach to describe the data information extracted from the pie chart in the form of natural language summaries. Kallimani and Srinivasa [30] designed a system to identify and interpret bar graphs and generate bar chart description from the extracted semantic information. Liu and Xie [31] proposed an approach to automatically generate chart description, and designed a summary template that can be extended to different chart types to generate a text description of chart.…”