Description
Our mission is to develop and test Natural Language Processing capabilities that can recognize, interpret, and characterize implicit levels of meaning in text without requiring human intervention, a substantial step forward in natural language understanding that would offer tremendous advantages to all its applications.
In order to accomplish this goal, we have undertaken a two-phased approach. The first is to conduct human subject experiments to better understand and model human ability to recognize the more subtle aspects of messages, and the second, based on these new understandings, is to apply Machine Learning algorithms to enable a system to accomplish this fully automatically.
The initial corpus of analysis is blogs, an unconventional, highly-personalized genre, and the initial application is Question-Answering, where our hypothesis is that the recognition of connotative meaning can contribute to richer / more subtle question-answering systems.
|
