Recognition and understanding of continuous speech
Automatic speech recognition

The use of isolated words considerably limits the richness of communication between man and machine. Only natural, continuous speech provides the level of expression required for applications of a certain complexity. This involves taking into account the process of encoding the speech signal before it is emitted by a speaker, and exploiting the information (acoustic, linguistic, etc.) required to understand the sentence or portion of sentence emitted. In addition to the context of the dialogue in which a sentence is inserted, the main sources of knowledge and information are as follows:

• acoustic-phonetic. This information governs the phonetic transcription of the message; to it must be added phonological information...