During his keynote speech at LILAC 2012, Honorary Visiting Fellow of Kellogg College, Lord Puttnam focused on the importance of oratory in education. He suggested both that being articulate, speaking confidently and understanding rhetoric are key to one’s success, and that voice recognition software is going to be the next big thing.
Commentators from the audience suggested that a child in a Glaswegian school might still struggle with voice recognition software. Lord Puttnam said he thought this would change soon by neccessity. It wasn’t quite clear whether he meant that the software would improve or that the accents would disappear.
I just happen to have a large collection of contemporary educational oratory -the Oxford Podcasts collection, which includes some fine examples of inspirational rhetoric and clearly communicated ideas. Our interactions with voice recognition software, however, have been frustrating.
It was always one aim of the OpenSpires project to offer transcripts to accompany our audio lectures. It is striking to note that even when both a video and audio version of a lecture is made available, the audio is downloaded by users three times more than the video version.
During the project the team explored various solutions including both automatic translation and human transcription services. Unfortunately the transcription service was so expensive as to be unsustainable for the volume of content we are publishing, and too slow given the turnaround time for which we aim. The automatic translators struggled with academic language, discipline specific terms, speakers’ accents, and could not pick up the subtleties of when a lecturer is quoting, making reference, or paraphrasing. The need to provide transcripts which present academic argument and do not confuse the reader regarding use of sources would be an issue dear to the hearts of all the LILAC delegates, I’m sure.
The need to provide transcripts is one dear to mine also as the Oxford podcasts service is often approached by external users citing our ‘responsibilities’ under the DDA.
In pursuit of reasonable adjustment, and because it is genuinely interesting, we have begun another project to explore how to best represent the content of our podcasts in text. By focusing on keywords generated by recognition software we will be able to give a searchable interface to users before they listen and represent the amount of relevant content within.