Background: Automatic Syllable Detection

Automatic syllable detection same as Vowel landmark detection?

• Most syllabic centers are vowels
  • syllabic sonorants appear vowel-like and act like vowels
• Most vowels are syllabic centers
  • elided vowel-vowel sequences ("coerce" -> "coarse")
  • vowel-semivowel sequences in different dialects ("fire" and "file")
• differences are open questions, rather than definite distinctions

History of ASD

• early work, mostly knowledge based
• recent work, mostly statistically based

Implications

• substantial literature on automatic syllable detection
• no consensus on how performance should be measured
• no publication addresses the ambiguities
• quoted performance numbers are difficult to compare to each other
• examples of knowledge based techniques combined with statistical models?