Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Integrating MAP and Linear Transformation for Language Model Adaptation

David Janiszek, Frédéric Béchet, Renato De Mori

LIA-CERI, Université d’Avignon, France

This paper discusses the integration of various language model (LM) adaptations. Ways of integrating Maximum A Posteriori (MAP) adaptation and linear transformation of bigram probability vectors are introduced and evaluated. This method leads to little improvements for adaptation corpora of less than 15,000 words. Another method, based on a data augmentation technique by means of a distance between history vectors in a reduced space is also proposed. This method allows us to improve the results when using adaptation corpora larger than 30,000 words.


Full Paper

Bibliographic reference.  Janiszek, David / Béchet, Frédéric / Mori, Renato De (2000): "Integrating MAP and linear transformation for language model adaptation", In ICSLP-2000, vol.2, 895-898.