Speech recognition and Natural Language Processing

APPROLING2 - Sebastien Bratieres

 

What speech and language engineering looks like

Natural language[1], just like human biology, has industrial applications and deserves the engineer's full attention. Moreover, language engineering tries to reproduce human intelligence, and therefore meets with the fascinating problems of artificial intelligence.

Arthur C. Clarke et Stanley Kubrick (2001, Space Odyssey) made an intelligent (and egoistical) computer, HAL, the main character of their story. We now write 2007, and HAL is still a long way down the road.

This course is on language understanding, and will describe the technical solutions that have been worked out over the years, from the enthusiastic beginnings of machine translation in 1955 to the latest achievements in Arabic/English speech-to-speech-translation sponsored by the US Army. This course, now is in its third year, offers limited places to allow for practicals.

Related disciplines are: artificial intelligence, computer science, signal processing, statistics and probability, linguistics, cognitive psychology.

How important it is becoming

·  The USA have about 55,000 call centres[2]. Speech recognition is now used now on speech-driven phone applications, thus partially automating some of the calls at a minimal cost.

·  The next generation of search engines is concept-enhanced. Google no longer makes it a secret that a large proportion of their R&D goes to linguistically informed machine learning techniques.

·  Natural language interfaces for domestic appliances, computers and cars is an obvious and irresistible trend. Sony puts speech recognition into its Playstation 3, Philips R&D works on speaking houses, Thales makes speaking car equipment, and so forth.

Cross-section of topics

Speech: phonetics, anatomy of speech production, speech signal processing. Speech recognition and synthesis, VoiceXML platforms. Pattern recognition, machine learning, speech and text mining.

Text: syntactic analysis (parsing). Statistical language modelling. Human-machine dialogue.

Lecturer

Sébastien Bratières (ECP 2001, MPhil in Computer Speech and Language Processing, University of Cambridge) is a speech and language engineer, currently working as a Senior R&D Engineer for Voice Insight. In the past, in startups or as a consultant, he developped speech-driven applications and virtual conversational agents. Since 2004, Sébastien has been mentoring études en autonomie or 2nd year projects on related subjects (linguistic mathematics, text-to-speech technology, machine learning), and another cours d'approfondissement in machine learning. Please contact him in case you would like to do an EA on an artificial intelligence topic, even if you can't attend this course.



[1]               As opposed to formal languages such as programming languages, the language of road maps, mathematical notation.

[2]               Reuters Business Insights, The Voice Business Market Outlook, December 2005.


Manager(s) for APPROLING2 : Sebastien Bratieres
Administrator for Les cours : TICE ECP
Powered by Claroline © 2001 - 2008