Recommendations on lexical surface information

1. Define the basic lexical entry type, as recommended in the previous section, and its notation.
2. Establish a machine-readable orthography convention.
3. Define the phonological level of representation (e.g. canonical phonemic).
4. Specify requirements in respect of word prosody  representation.
5. Specify requirements in respect of phrasal and discourse prosody  (intonation ) information.
6. Use the European standard machine readable phonemic alphabet SAMPA (see Appendix B).
7. Select an appropriate system of prosodic  representation, such as SAMPA , SAMPROSA (see Appendix B), INTSINT, or ToBI  notation, where required.
8. Ensure that the relation between notations and representations used in the lexical database  and the system lexicon are well-defined, and that they are completely consistent with notations and representations in other resources such as corpora and in the different parts of the system, such as the word lattice  produced by the speech recogniser , operated on by a stochastic language model , and further processed by a sentence parser.