Publications

Huenerfauth, Matthew Paul. The development of PETALS: Personality Tagged Logical Statistical Generator. Master's Thesis, Department of Computer and Information Sciences, University of Delaware, Newark, Delaware, Spring 2001.

ABSTRACT:

The subject of this thesis is the development of the PeTaLS generation architecture designed to work with the ICICLE English-writing tutorial system. The Personality Tagged Logical Statistical Generator uses machine learning and logical rule based techniques to accomplish the deep generation (content selection and propositional ordering) of the instructional text to be delivered to the student. Training on a corpus of text written by English as a Second Language instructors, the system can learn to mimic the informational organization strategy of individual authors. This thesis includes a comparison between PeTaLS and other deep generation architectures, the advantages of machine learning to generation, training data collection and analysis, and a discussion of system development considerations.

The ICICLE Project • Department of Computer and Information Sciences • University of Delware• 2007