Graduate School of Language Technology in Finland

Kieliteknologian valtakunnallinen tutkijakoulu - Språkteknologiska forskarskolan i Finland


Course in Emergent Phonology (1 - 2 cr)

Lecturer

Björn Lindblom (Stockholm University & University of Texas at Austin)

Location

Department of General Linguistics
Siltavuorenpenger 20 A
Helsinki

Course description

The goal of this course is to paint a broad picture of current research on speech processes including adult production and perception of speech as well as learning to speak and acquiring one's native phonology. A significant difficulty for all concerned with research in those areas is how to handle the variability of the speech signal, a classic problem known as the 'invariance issue' (the one-to-many correspondence between abstract linguistic categories and their noisy, incomplete and variable manifestations in the signal). It is made particularly evident it in the context of normal language acquisition: If the signals that they hear during development are so variable, how are children able to find the underlying structure, the phonological entities and rules, of their native language?

In the first half of the lectures we will provide a comprehensive review of the literature on how on-line adult speech communication works. In the second half we will move into phonology and attempt to address how children go about resolving the invariance issue. This leads us to Emergent Phonology (EP), a paradigm that I and my colleagues have developed over the years. The EP agenda springs from the following question: If sound systems were seen as evolutionary adaptations to universal conditions of language use - constraints of speaking, listening and learning to speak, ?. - what would they be like? You will hear what EP is, why it is needed, how it differs from other approaches, how you go about doing it, what has been accomplished so far and what the research needs are. EP views phonological units as emergents of phonetic learning. It uses simulations, emergent computation and models of unsupervised learning to explore that hypothesis. Consequently, it differs radically from the 'formal' linguistic approaches to phonology in insisting that structure (entities and rules), but be derived from behavioral principles independent of the data to be explained.

In addition to lectures and plenty of discussions (I hope!) we will also spend some time doing small computational projects designed to illustrate the issues reviewed in class.

Duration and Organization

The course will take place as follows:

Dates Lecture themes Seminar Lab
17 - 18 May How speech works: An Overview
Review of current theories of speech production and speech perception
Adaptive organization of speech: Resolving the invariance issue
debates and discussion project time
19 - 20 May Speech development: From signal to phonology
Introducing Emergent Phonology
debates and discussion project time

Back to the Courses Page

up


Last updated: Tuesday, 27-Jan-2004 13:04:10 EET