Publication at Faculty of Mathematics and Physics |
2015
Abstract
A dialog state tracker is an important component in modern spoken dialog systems. We present the first trainable incremental dialog state tracker that directly uses automatic speech recognition hypotheses to track the state.
It is based on a long short-term memory recurrent neural network, and it is fully trainable from annotated data. The tracker achieves promissing performance on the Method and Requested tracking sub-tasks in DSTC2.