Time: 2pm-3pm, Friday, Dec 10, 2010
Place: Room 4102, CUNY Graduate Center, 365 Fifth Ave (34str&35str).
Speaker: Fadi Biadsy (Columbia)
Title: Accent and Dialect Recognition

Abstract:

Accent and dialect recognition have recently begun to receive attention from
the speech science and technology communities. Dialect recognition helps in
Automatic Speech Recognition (ASR) since speakers with different dialects
pronounce some words differently, consistently altering certain phones and
even morphemes. In this talk, we will describe a system that can detect the
regional dialect and native accent of the speaker given a sample of his/her
speech. We will first describe and examine  previously explored approaches
employed in the language recognition community. We will introduce our novel
approaches: Discriminative Phonotactics and Kernel-based methods which rely
on the hypothesis that certain context-dependent phones are realized
differently across dialects. Our system significantly outperforms previous
approaches. The system we have built can also be used to extract linguistic
knowledge, particularly subtle phonetic differences that contribute to
distinguishing dialects.

Speaker Bio:

Fadi Biadsy is a PhD candidate in the computer science department at
Columbia University, working with Prof. Julia Hirschberg. His research
interest is spoken language processing and natural language processing. His
dissertation work focuses on automatic spoken dialect and accent recognition
and its application for automatic speech recognition. He has also explored
various other topics, including charismatic speech, question answering,
extraction of biographical information, sentiment analysis, and online
handwriting recognition. Before arriving to Columbia, he received a BSc and
MSc in computer science from Ben-Gurion University. He also worked as a
software engineer at Dalet Digital Media Systems.