RONI ROSENFELD
Professor, Language Technologies, Automated Learning and Discovery and Computer Science
www

My research interests span the following:

Computational molecular virology and vaccine design.   Retroviruses like HIV and RNA viruses like Influenza evolve at a much higher rate than DNA life forms.  This is a formidable challenge to vaccine design, but is also an opportunity to observe evolution as it happens.  We use the fast growing databases of viral sequences to build descriptive and generative models of viral molecular evolution.  We also use them to infer viral envelope properties and suggest potential antigenic targets that cannot easily mutate away.  In collaboration with virologists/immunologists, we try  to correlate isolate sequence composition to important biological properties of the isolate, such as pathogenicity, infectivity and neutralizability.  Along they way we design and develop visualization tools for multiple sequence alignments (MSAs) and other biological sequence data.

Computational Molecular Biology and more specifically Computational Biolinguistics. Many of the problems in this area involve statistical modeling of long sequences of symbols/building blocks (nucleotides or amino acids) and their relationship to proteins and their function. This is very similar to the problem of modeling natural language: long sequences of symbols (letters, words), and their relationship to the deep structure and meaning of sentences. We are hoping that some of the models and techniques we have developed in the past decade for language modeling will prove useful in the biological domain.  Current projects include computational molecular evolution, computational virology, and multi-species gene-expression analysis.

Speech and Language Technology for Development (SLT4D) is the term we coined for our own subfield of ICT4D: finding ways to use speech and language technologies (like automatic speech recognition and human-machine dialog systems) to help people around the world help themselves.  Our current project, HealthLine, investigates the use of spoken language interfaces for community health workers across Pakistan (thank you, Microsoft!).

The Future of Human-Machine Speech Communication.   Communication with machines and information-servers does not require the full strength of natural language, nor should it have to cope with its ambiguities. What then is the ideal form of human-machine speech communication? Will there develop a particular style for talking to machines?  If so, can we help this process along by developing principles for it?  In the Universal Speech Interface (USI, a.k.a. "Speech Graffiti") project, we develop and test such principles.  In essence, we are trying to do for speech communication what Graffiti™has done for mobile text entry (see also The USI Manifesto).

Statistical Language Modeling for Human Language Technology applications, such as automatic speech recognition, machine translation, topic detection, information retrieval and textual data mining (not looking for new students in this area). 

Personal web site:  http://www.cs.cmu.edu/~roni