Linguistic Expression of Emotion in Human-Machine Interaction: The NIMITEK Corpus as a Research Tool
In: Proceedings of the Fourth International Workshop on Human-Computer Conversation (Bellagio 2008). The Companions consortium, Bellagio, Italy.
Since the end of 2005, the NIMITEK consortium investigates issues in spoken human-machine interaction (HMI). We employ the multimodal NIMITEK corpus of affected behaviour in HMI as a tool that provides an empirical foundation for the development of integrated emotion detection from both signal based as well as content based emotion recognition. This paper primarily discusses various linguistic features that may carry affect information (e.g., key words and phrases, lexical cohesive agencies, dialogue act sequences, etc.). Finally, we report on a first prototype of an automatic annotator for recognition and tracking of the user’s emotional state from linguistic information.
(6 pages, no pagination)