Phoneme Inventory in Corpus for Speech Synthesis
|Submitter Email:||click here to access email|
I am currently experimenting with a TTS system in General American and researching my dissertation on improving speech synthesis and TTS systems with phonological/phonetic knowledge. I have generated a series of phonological rules in order to more accurately label the the corpus by using alternatives in the segmentation. I am only dealing with within-word processes for the moment. The dictionary used to transcribe the corpus employs the symbols used in LPD notation (phonemic). It does not include the symbols of narrow phonetic transcription (allophones eg. aspirated voiceless stops, tap-flaps etc.).
In my rules i have integrated new symbolic symbols for tap-flaps and syllabic consonants. As I am considering including a rule related to aspirated voiceless stops, I am wondering if most English speech synthesis systems include these sounds as separate symbols in the transcription. That is, I wonder if most commercial systems differentiate between aspirated and non-aspirated voiceless stops at the transcription level.
The transcription is based on a modified version of the Festival Dictionary 0.4. As it is difficult to find any other freely available dictionaries for synthesis, I would greatly appreciate knowing what level of transcription they generally use or an inventory of symbols commonly used.
Please contact: firstname.lastname@example.org
Sums main page