next up previous contents index
Next: Channel adaptation/Environment adaptation Up: Speech recognition systems Previous: Speaker selection

Language modelling


The language model is used to improve the performance. Different paradigms can be used: finite state syntax,  external NLP module    with a parsing  grammar , an n-gram  (unigram , bigram , trigram ) model, etc.

The language modelling (see Chapter 7) may be tuned by the technology provider with respect to a particular variety of languages. An adaptation to a specific application can help to improve the results, and may or may not be necessary. This tuning may be done by the technology provider or by the application developer. A tool or a formalism that allows data to be entered may be provided and any data that has to be collected by the application developer should be indicated.  

EAGLES SWLG SoftEdition, May 1997. Get the book...