FAST stands for Foreign Accented Speech Transcription and is being developed by Rio Akasaka, a senior Linguistics and Engineering double major at Swarthmore College. The goal of the game is to provide a more intuitive interface for human recognition of foreign accented speech, as well as to facilitate the transcription of large corpora.
There are two different types of questions you will be asked in this game. In the first, you'll be asked to identify the native language of a foreign accented speaker of English. In the second, you'll be asked to write down some of the words you hear in a recording.
This game serves as a collaborative information retrieval tool that will enable easier and less tedious transcriptions of foreign accented speech. Existing systems have considerable difficulty transcribing foreign accented speech of poor quality or varying severity of accent. By proving an entertaining and practical tool that can be deployed on a social networking site like Facebook, the hope is to provide another method in improving the language models that computers currently use in transcribing foreign accented speech.
See the privacy statement