|
Advisor: Dr. Alexander Rudnicky CMU Communicator Project CMU Speech Group Stochastic NLG
|
More and more, people are
becoming interested in interfaces other than GUI and input modality other
than typing. Speech is a natural choice.
Speech obviously cannot be the only way to communicate with machines,
but it is a natural, convenient,
and efficient way in many situations.
In the CMU Communicator project, our goal is to enable a natural conversation between a human user and a computer travel agent. This involves the use of latest language technologies such as the Sphinx speech recognizer, Phoenix parser, agenda-based dialog manager, and stochastic natural language generation.
Natural Language Generation (NLG) is the process of making machines talk or write in natural language. Even though people learn to talk well before they learn to write, considerably less research effort has been put into making machines talk than making machine write. Also, much more research has been done in natural language understanding (NLU) than in NLG. My research goal has been twofold: to highlight the importance and non-triviality of generating spoken language, and to explore corpus-based techniques for natural language generation. |
