My main reseasrch interests focus on spoken language processing; particularly, spoken dialect recognition, automatic speech recognition, and identifying charismatic speech.
Dialect Recognition:
In the past years considerable attention has been paid to the automatic
recognition of languages using acoustic-prosodic information alone.
That is, how can we identify the language a speaker is speaking from the acoustic signal alone?
In recent years, the identification of accents and regional dialects has attracted interest
from the speech community. Can a speaker’s regional origin or regional dialect within a given language group be determined
given a small sample of his or her speech? Our goal is to recognize the dialect of a speaker in aid of improving Automatic Speech Recognition (ASR). Since speakers with different dialects often pronounce
some words differently, consistently altering certain phones and
even morphemes, identifying regional dialect prior to ASR allows
for the use of a more restricted pronunciation dictionary
in decoding, resulting in a reduced search space and lower perplexity.
Charismatic Speech:
Automatic analysis of a speakers' charisma can serve to diagnose or improve presentational skills as well as (potentially),
to identify new political leaders, or to improve speech synthesis when compelling speech is needed. We identify acoustic, prosodic, and lexical features that correlate with charisma and compare perceptions of charisma across cultures (American English, Palestinian Arabic and Swedish)
Biography Generation:
I've also been working on a multi-lingual and multi-media question answering system, part of DARPA Gale porject, particularly on automatically generating biographies. I use Information Extraction (IE) and text summarization techniques.
Previosuly, I explored the problem of Online Handwriting Recognition of Arabic script. I'm still interested in conducting more research in this problem, hopefully sometime after completing my PhD. See below a screenshot of the system I developed.
