WordNet Visualization
Christopher Collins and Gerald Penn

Abstract
Interface designs for lexical databases in NLP have suffered from not following design principles developed in the information visualization research community. We present a design paradigm and show it can be used to generate visualizations which maximize the usability and utility ofWordNet. The techniques can be generally applied to other lexical databases used in NLP research.

A short paper on this project is available in PDF.

Despite the growing dependence on statistical methods, many NLP techniques still rely heavily on human-constructed lexical resources such as WordNet. While development onWordNet continues, the interfaces for interacting with WordNet have not progressed to take advantage
of advances in the field of information visualization. Currently available interfaces, both textual and graphical, focus on regions of local interest, for example by searching for the relationships for a single synset. In this work, we follow a well-accepted design paradigm to create a working prototype of a visualization suite for WordNet which allows for an overview of the data, as well as the ability to focus on specific synsets of interest and obtain details. For full details, please refer to our short paper.

Hyponymy of “emotion” filtered at depth 2. Words starting with “h” and collapsed nodes containing search results are highlighted in pink. (B) Graph from (A) changed so that subtree rooted at the synset containing “love” is the central focus node, expanding its radial extent. (C) All highlighted search results from (A) expanded.

This work was created with the excellent prefuse information visualization toolkit.