The material presented in this chapter is relevant both to corpus characterisation and to design and evaluation issues in automatic speech recognition. The chapter is different from most of the other chapters:

- Language models as such are an important part of the whole recognition system , and therefore this chapter will give more details of the techniques and algorithms.
- To fully understand the problems and implications of language modelling, it is necessary to consider some of the statistical and mathematical problems involved.
- The sections are written in such a way that, at first reading, the reader may skip the mathematical parts.
- Most of the mathematical details and derivations have been put in an appendix (see 7.8).

The reader should consult also the following chapters:

- spoken language lexica,
- dialogue,
- assessment methodologies and experimental design,
- assessment of recognition systems .

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