A Perspective on Relevance and Application of Prosodic Information to Automatic Recognition in Italian

In a free-stress language like Italian, automatic detection of stressed syllables is of great importance for access to the lexicon. In continuous speech recognition systems this information can facilitate lexical access since it constrains the location of word boundaries and can reduce the process of word hypothesizing and matching. The paper summarizes the results of acoustic and perceptual studies on prosodic features related to lexical stress and outlines a relatively simple algorithm based mainly on vowel duration for automatic detection of stressed syllables.