Towards an Annotation Standard for STEM Documents

Abstract

When publishing papers, researchers in mathematics and related disciplines typically focus on the presentation, i.e. type-setting, of their ideas and provide little semantic information. This impedes the development of services that benefit from semantic information, such as semantic search and screen readers for vision-impaired researchers. As a remedy, there have been attempts to infer semantic data from already published papers using small programs that we call spotters. Unfortunately, there is no standardized format for semantic annotations and spotter authors typically invent their own format. This leads to two problems: i) there is no ecosystem of tools for common tasks like the visualization of results or the manual annotation of a gold standard, and ii) re-using, evaluating and combining results becomes very difficult.

In this paper, we address these issues by describing a standardized, flexible way to represent semantic annotations, using semantic web technologies and, in particular, the Web Annotation standard. Furthermore, we describe SpotterBase, a set of tools to help with processing the annotations and creating new ones.