9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Automatic Detection of the Context of Acoustic Landmark Deletion

Nanette Veilleux (1), Stefanie Shattuck-Hufnagel (2)

(1) Simmons College, USA; (2) MIT, USA

Earlier work has shown that the acoustic landmarks in speech [1,2], proposed to be important in lexical access, are largely preserved in spontaneous American English Speech [3]. Moreover, the loss of acoustic landmarks predicted from word's lexical representation is systematic and predictable. This study reports preliminary analysis of factors that govern whether a landmark will be realized as predicted, or will be changed or apparently omitted. A classification tree is designed to predict the deletion/change or preservation of a landmark in a corpus of spontaneous American English using contextual factors that include prosody, word structure, morphosyntactic categories and landmark type.


  1. Stevens, K. Acoustic Phonetics. Cambridge, MA: MIT Press, 1998.
  2. Stevens, K.N. "Toward a model for lexical access based on acoustic landmarks and distinctive features". J.Acoust.Soc. Amer. 111, 1872-1891, 2002
  3. The Robustness of Acoustic Landmarks in Spontaneous Speech", ICPhS, Saarbrücken, 2007

Full Paper

Bibliographic reference.  Veilleux, Nanette / Shattuck-Hufnagel, Stefanie (2008): "Automatic detection of the context of acoustic landmark deletion", In INTERSPEECH-2008, 2847-2850.