Intermediate level speech processing as an aim of low level signal analysis

In the general area of speech recognition, two processing levels have dominated recent literature: low-level signal processing aimed at segmenting and spectrally analysing speech input; and high-level syntactic and semantic processing, part of natural language processing (NLP). Both areas cover vital aspects of speech, but they do not readily link together. Greater activity is needed to link these two into complete systems. The linkage of the two is taken to define an intermediate speech processing level. The authors examine the processes which constitute this interface (they are disambiguation processes), reporting on the phenome lattice as the end-product of the signal processing level