The main aim of the project is to develop a system which could automatically recognise aggression and potential threats in Hindi and Indian English speech. Based on the theoretical research in aggression and (im)politeness, we will develop the prototype of a system which would be able to recognise elements of aggression and impending physical violence as indicated by several phonological cues like the prosody of the utterance and the kind of lexical items used in the utterance.
The most important practical outcome of the project will be the development of two very important resources – a multimodal corpus of Hindi and Indian English and an aggression detection system. The corpus could be very useful for several purposes including the theoretical studies in linguistics as well as other computational linguistics as well as artificial intelligence tasks. The corpus collected during the project will definitely prove to be one of the most long-lasting and useful outcomes of this project. This corpus will be of aggressive behaviour and it will be annotated for the presence of real aggression, which makes it much more rich and useful than a raw corpus, particularly for such applications as sentiment analysis.
In addition to the corpus, the prime objective of the project is to develop an aggression detection system, which could prove to be very useful in crime detection and control. This system could also be used for further research in aggression and also inter-cultural understanding and communication.
Finally several other smaller tools and system will be developed during the project for jobs ranging from collection, annotation and arrangement of speech data to systems for facilitating the collaboration by providing better ways of monitoring as well as sharing information and resources over the web. These tools, like the corpus and the main tool, will definitely prove to be useful beyond this project itself.