THESIS: Reference Specification in Multilingual Document Production

Written documentation must be provided to people around the world who speak many languages. Traditionally, two approaches have been used to arrive at the language-independent representation necessary for generating documents in multiple languages: direct specification by a person or automatic extraction from a natural language document. Recently, research has focused on an approach lying in between these extremes in which a person interacts with a text-based user interface to construct the language-independent representation. Like natural language, one of the problems inherent in this approach, known as symbolic authoring, is the problem of reference.

This thesis presents methods and provides empirical evidence demonstrating how to make both the specification of entities in the user interface for knowledge-base editing and the generation of expressions to refer to these entities in documents more natural. More specifically, this thesis describes the development of three types of reference mechanisms: 1) a statistical model that uses domain and lexical knowledge to organize new semantic options in the interface; 2) techniques for controlling coreference specification that take advantage of discourse and task structure; and 3) learned models for generating expressions to refer to new and already mentioned entities contained in the specified knowledge base. The evaluation of these reference mechanisms demonstrates the following: 1) specifying new entities using an interface informed by computational linguistic processing is more helpful than traditional knowledge-editing interfaces; 2) exploiting discourse and task structure reduces the amount of time required for people to refer to entities in the interface that have already been specified in the knowledge base; and 3) using learned linguistic information to generate referring expressions in documents leads to expressions that more closely match the decisions of people. [PDF]

PAST PROJECTS

Harvard University, 1999-2001:  Intonation and Discourse Structure
Advisor:  Professor Barbara Grosz

Used read and spontaneous speech to automatically identify elements of discourse structure based on intonational features.  The learned models provide further insight into how speakers use acoustic-prosodic properties to convey information about the structure of discourse. [PDF]

Lucent Technologies, Summer Intern 1999:  Dialogue Systems
Advisor:  Dr. Jennifer Chu-Carroll

Conducted experiments to evaluate the mixed initiative and automatic adaptation aspects of an adaptive mixed initiative spoken dialogue system.  Analyzed the resulting dialogues along three dimensions:  performance factors, discourse features, and initiative distribution. [PDF]

Lucent Technologies, Summer Intern 1998:  Intention Recognition
Advisor:  Dr. Jennifer Chu-Carroll

Worked on a project involving the use of acoustic-prosodic features to disambiguate direct and indirect speech act productions.  Used xwaves speech analysis software. [PDF]

PUBLICATIONS

Linguistic Coreference in Knowledge Editing

• Nickerson, J., Shieber, S., and Grosz, B. 2006.  Referring-Expression Generation Using a Transformation-Based Learning Approach.  In Proceedings of the 19^th International FLAIRS Conference, Melbourne Beach, Florida.
RECIPIENT OF BEST PAPER AWARD
• Nickerson, J.  2005.  Reference Specification in Multilingual Document Production.  Ph.D. Dissertation, Harvard University. Harvard CS Technical Report TR-21-05.
• Nickerson, J.  2003.  Statistical Models for Organizing Semantic Options in Knowledge Editing Interfaces.  In Proceedings of AAAI Spring Symposium Series Workshop on Natural Language Generation in Spoken and Written Dialogue, Palo Alto, California.
• Nickerson, J.  2002.  Controlling Linguistic Coreference in Graphical Interfaces.  Student Research Workshop Paper.  In Proceedings of the 40^th Conference of the Association for Computational Linguistics, Philadelphia, Pennsylvania.

• Nickerson, J. and Chu-Carroll, J.  2000. Evaluating Automatic Dialogue Strategy Adaptation for a Spoken Dialogue System.  In Proceedings of the 1^st Conference of the North American Chapter of the Association for Computational Linguistics, Seattle, Washington.

• Nickerson, J. and Chu-Carroll, J.  1999. Acoustic-Prosodic Disambiguation of Direct and Indirect Speech Acts.  In Proceedings of the 14^th International Congress of Phonetic Sciences, San Francisco, California.

Linguistic Coreference in Knowlege Editing

• At IJCAI (International Joint Conference on Artificial Intelligence) 2003 in Acapulco, Mexico, I presented my thesis proposal in the area of linguistic coreference in knowledge editing at the SIGART/AAAI/IJCAI Doctoral Consortium.

• At Grace Hopper Celebration of Women in Computing 2002 in Vancouver, BC, Canada, I presented a poster describing my work in the area of linguistic coreference in knowledge editing.

• For my qualifying exam, I wrote a paper on using acoustic-prosodic features to automatically identify elements of discourse structure.

TEACHING EXPERIENCE

Teaching Fellow in Spring 2003:

              CS 288 - Computational Models of Discourse

Teaching Fellow in Fall 2000, 2001, and 2002:

              CS 182 - Intelligent Machines:  Reasoning, Actions, and Plans

GUEST LECTURES

Guest Lectures on Discourse in Spring 2004:

              CS 187 - Computational Linguistics

CONTACT INFORMATION

857.928.8399 (voice)
702.924.5008 (fax)

jickerso521 at yahoo dot com