Domain Specific Intelligent Personal Assistant with Bilingual Voice Command Processing

Intelligent Personal Assistants (IPA), like Siri and Alexa, are created to assist their users with simple digital tasks. Here, we propose the steps we have used to develop a voice operated IPA which can process direct commands in two languages: English and Bengali, to perform menial tasks for the users. The speech recognition engine of the IPA is constructed with Sphinx-4, and the language processing is performed by a modified finite state automaton. The IPA also takes advantage of the subject/action structure of commands to reduce the size of the word domain, and utilizes a generalization function to ensure that the language processor can understand multiple languages without undergoing major modification - making this approach suitable when training data is limited.