GailBot: An automatic transcription system for Conversation Analysis
GailBot could probably be run on Red Hen data. Would you like to try?
If so, write to
and we will try to connect you with a mentor.
- The basic article on GailBot is at https://osf.io/ap4s7/. Umair, Muhammad, Julia Mertens et al. Under Review. "GailBot: An automatic transcription system for Conversation Analysis."
- "To summarize, GailBot is the first automatic transcription system to use Jeffersonian notation to represent turn-taking, pauses, gaps, overlaps and laughter. While each of the modules have limitations, GailBot represents a substantial improvement over previous systems. Further, we purposefully designed GailBot in a way that would facilitate future development. As researchers learn more about how interlocutors and transcribers identify changes in pitch, volume,and other features of conversation, new modules can be changed or added. For example, there is no consensus about how sensitive professional transcribers are to context when it comes to relative markings, such as changes in speech rate or volume. Do Conversation Analysts mark segments that are faster or louder than the previous minute, ten minutes, or ten turns by the same speaker? Similarly, incremental improvements in individual parts of the system, such as the quality of STT services, or the accuracy of our laughter detection algorithm will steadily enhance key aspects of GailBot’s performance."