AAIA’17 Data Mining Competition
Automatic Transcription of Spoken English
The objective of the competition will be to create a method of automatic transcription of spoken English that adopts a combination of audio and visual modalities, which provide a synergy effect in terms of speech recognition accuracy. The approach is assumed to be sensitive to minute allophonic details as well as to accentual differences. The data will consist of more than 600 words pronounced by several native and non-native English speakers. Words will be represented by various features extracted from the audio signals and from accompanying video images of the mouth. In this way, we aim to verify whether by analyzing video together with acoustic signals we can facilitate the robust speech transcription.
The competition awards will be the same as at the previous FedCSIS conferences.
The competition web page will soon be available at the Knowledge Pit platform.
On the behalf of Competition Committee, we would like to cordially invite you to participate.
Andrzej Czyzewski & Andrzej Janusz - Competition Chairs