Skip to Main Content
Syllabic structures in English are rather complex and have not shown promising results for continuous speech recognition (CSR). However, in the Portuguese language these structures are simple and constitute the nucleus on which words are formed. The use of syllables as subword units for Portuguese CSR systems is investigated. Another aspect investigated concerns the role of function-words in CSR algorithms. In English they are responsible for a large percentage of word recognition errors. The importance of these words in CSR systems for Portuguese is also investigated.