Marty Meinardi, from the Dublin Institute of Technology, talked about a very important issue, namely the [English] language learners' difficulties in finding a match between the sound they hear and the words they know. When native speakers of English speak very fast, the speech becomes messy, and this makes understanding difficult for the learners Special attention should be paid to this in teaching.
Marty quoted Swan, who has written that "if students are exposed only to scripted material, they will learn an impoverished version of the language, and will find it hard to come to terms with genuine discourse when they are exposed to it". Listening exercises are often read aloud from text by actors and are not authentic. So there is a great demand for "real" language to be used in listening exercises.
There have been programs that can slow down sound files, but they have been created for music purposes and mostly don't give good results with speech. A new algorithm has been developed at the DIT for slowing down speech. It looks at the sound pattern and can slow down speech down to 40 % of the original time without the speech sounding too odd (although it sounds a bit sleepy). But at 80 %, even the native speakers consider the quality very good and natural.
According to Marty, the slow-down tool can be used to facilitate the processing and retrieving formulaic sequences and chunks which the native speakers usually speak at a faster rate. This way these highly important elements of fluent speech become more accessible and "learnable" to the students.