Commercial spoken dialogue systems traditionally have been static in the sense that once deployed, these applications only get updated as part of formal releases. Also, the creation of classification grammars in natural language call routing applications requires expensive manual annotation of caller intents. The work presented here introduces a process to semi-automatically annotate new data and to use the new annotations to update the training corpus to continually improve the classification performance. This new method consists of a combination of using multiple classifiers in a voting schema to automatically classify an utterance and a boosting mechanism to continually update the classifier with the new automatically annotated training data. This method was tested with 6 weeks' worth of data from a live system. It is shown that with this approach about 94% of all new utterances can be automatically annotated. Using the iterative boosting approach increased the size of the training corpus by about 6% per iteration while at the same time slightly increasing the classification accuracy.
Bibliographic reference. Witt, Silke M. (2011): "Semi-automated classifier adaptation for natural language call routing", In INTERSPEECH-2011, 1341-1344.