Location

MIT Media Lab, E14-633

Description

In general, an ontology is a set of concepts and relations between those concepts. Since ancient times, ontologies have been used as a means to organize and access information. Most ontologies are large and complex. This is because they aim to cover a wide range of topics and support a large set of users and tasks. Some situations, however, call for more lightweight ontologies that are user- and task-specific. For example, in lawsuits and regulatory reforms, lawyers and government employees must quickly organize large amounts of materials into task-specific concept hierarchies that will later be discarded. Complex ontologies are not well-suited for these situations. This talk examines “personal ontology learning,” a mechanism to create lightweight and personalized ontologies that allow users to quickly understand the range of the issues raised, and enable “drilling down” into documents that discuss a specific topic.
In this talk, Yang will highlight her work on personal ontology learning. In the first half, she will illustrate how to optimize the ontology structure, and how to model concept abstractness and long-distance relations. In the second half, she will present how to leverage machine learning to produce a user- and task-specific ontology interactively. She will also present her findings on user behaviors during ontology construction. Example findings include whether there are consistent dataset-specific or user-specific differences in the ontologies that people construct, whether people are self-consistent, and how these factors interact with the construction methods.

Biographies

Grace Hui Yang is a PhD candidate at the Language Technologies Institute within the School of Computer Science at Carnegie Mellon University. She received her master’s degree in computer science from the School of Computer Science at Carnegie Mellon University, and bachelor's degree in computer science from the School of Computing at the National University of Singapore. Yang’s research interests lie at the intersection of information retrieval, text mining, machine learning, and natural language processing, with a recent extension to human-computer interaction. Her current research includes automated and interactive ontology generation, human-guided machine learning, and text analysis and organization. Prior to this, she conducted research on question answering, near-duplicate detection, search engine training and evaluation, multimedia information retrieval, and opinion and sentiment detection. Her work on question answering generated the award-winning Question Answering (QA) system in the Text REtrieval Conference (TREC) evaluations from 2002 to 2004. Her work on near-duplicate detection has been used by three US government agencies for fifteen regulatory activities, and yielded a spinoff company. Her intern work on search-engine training and evaluation is currently employed by Microsoft Bing. She has published more than 20 research papers at conferences such as SIGIR, ACL, WWW, ACM Multimedia, CIKM, TREC, TRECVID, EACL, COLING, HCIR, and DG.O.