Resources
From SpeechWiki
Contents |
Transcription tools
Scripts
- To get started, transcribers need to download and install the Praat interface scripts.
- (Note: For archival purposes only, previous version of scripts is here).
- While transcribing, transcribers can use the script remove to delete a selected boundary as well as all other boundaries that are perfectly time-aligned across feature tiers (tiers 2 - 10).
- After finishing transcribing the whole utterance of a file, transcribers should run the script check_correctness to check if there are any mistakes in their transcripts. It checks the following three types of errors:
- Is there any "*" (unspecified features)?
- Is there any empty interval?
- Any violation to the vowel rule? i.e., There is "NONE" on the tier 2, but there is no vowel label in the tier 10. ("hh" is an exception to this rule, and it is taken care of by checking the tier 8 for "ASP".
Data
- New transcribers should practice the transcription guidelines by labeling these practice utterances.
- UIUC site, official transcription utterances
- Set 1 (15 STP utterances).
Completed transcriptions
- TTI Set 1 (Katie & Matt, both before and after discussion -- missing one of Katie's files).
- UIUC Set 1 (Heejin & Mindi, both before and after discussion).
References
- Manual transcription of conversational speech at the articulatory feature level, Karen Livescu et al., ICASSP, 2007.
