EasyTree: A Graphical Tool for Dependency Tree Annotation

EasyTree: A Graphical Tool for Dependency Tree Annotation Alexa Little∗, Stephen Tratz† ∗Yale University, New Haven, Connecticut, USA †U.S. Army Research Laboratory, Adelphi, Maryland, USA alexa.little@yale.edu, stephen.c.tratz.civ@mail.mil Abstract This paper introduces EasyTree, a dynamic graphical tool for dependency tree annotation. With EasyTree, annotators can incrementally construct and label trees by manipulating an interactive graphical tree representation and then exporting the internal data representation as JSON (JavaScript Object Notation). This intuitive animated interface has a number of customizable features to assist annotators, including color-coded part-of-speech indicators and optional translation displays. It can be adapted to suit a range of different projects. Edge labels, part-of-speech categories, and many other settings can be edited from within the GUI. EasyTree uses the standard UTF-8 encoding internally and properly handles both left-to-right and right-to-left scripts. In providing a user-friendly annotation tool, our aim is to significantly reduce the time spent transforming data or learning to use software, to improve the overall user experience for annotators, and to make annotation approachable even for inexperienced users. EasyTree is built entirely with standard web technologies—JavaScript, HTML, and CSS. Thus, it is well-suited for web-based annotation efforts such as crowdsourcing efforts. Keywords: annotation tools, dependency trees, visualization, GUI 1. Introduction Manually constructed linguistic resources are crucial for the training and evaluation of most state-of-the-art natural language processing tools, including dependency parsers, and enable countless research efforts. Unfortunately, the annotation process is typically very slow and tedious. This is particularly true for more complicated tasks such as de- pendency tree annotation. Although some tree-editing pro- grams have been developed to address this issue, most ei- ther rely on text-based editing, which can be slow and error- prone, or are standalone applications designed for sophisti- cated users, assuming a moderate knowledge of linguistics or computer programming. These issues are inconvenient for researchers and annotators generally but are especially problematic for low-resource language tasks, for which ex- perienced annotators may not be available, and crowdsourc- ing efforts, which typically require a browser-based anno- tation tool. To address these needs, this paper introduces EasyTree,1 a dynamic visual tree editing system designed to make tree annotation fast and intuitive. To reduce the time and cost of training annotators— significant issues, especially for low-resource language an- notation tasks—EasyTree provides a clean, uncluttered in- terface. It provides important functionality for dealing with large trees such as panning, zooming, and subtree hiding, and it relies on intuitive, well-accepted metaphors such as drag-and-drop. To ease deployment, EasyTree is built entirely with JavaScript, HTML, and CSS—standard web technologies that are supported by all major web browsers. Thus, EasyTree is easily deployed within web browsers, making it appropriate for online web-based annotation tasks such as crowdsourcing efforts. If running on a remote server, no client side installation is necessary; otherwise, installation is as simple as decompressing a .zip file. In the remainder of this paper, we give an overview of EasyTree’s use and important features (Section 2), discuss 1Available at https://github.com/alexalittle/easytree the implementation of EasyTree (Section 3), and compare and …

Content truncated to 500 words