Statistical Speech Technology

Renoir's Luncheon of the Boating Party depicts six simultaneous conversations in a small outdoor restaurant
Renoir: Luncheon of the Boating Party

The Statistical Speech Technology group was created in 1999 at the University of Illinois. Our research is motivated by poetry, mathematics and empowerment.

Poetry, rhyming, song, hip-hop and swagger are a riotous fern jungle shaped to the beautiful forms of human speech. Speech is beautiful because it is the perfect balance between the human needs to communicate information and to express social belonging, expressed and filtered by the tongue and the air and the ear.

Mathematics is the art of expressing truth as precisely as possible. Information-theoretic models of speech make it possible to diagnose and treat human disabilities and translate text. Signal processing models permit us to separate the linguistic and nonlinguistic components of speech, understand them, synthesize them, and bring computers into the conversation.

Both justice and progress are best served by a society in which every person is empowered to learn, love, and create. Communities form around modes of speaking, and speech is therefore a useful technological tool with which to empower communities. Disability is a failure of technology: working technology enables communication, collaboration, and development.

We are people who work together on projects and publications, create datasets, teach courses, and speak to the press. We sometimes have parties (this activity has become more frequent since the end of the pandemic).