| 0:00:15 | i apologise we don't have a slight proposed ten but the posters on identification approach |
|---|
| 0:00:20 | regional accents using fusion of i-vector and multi-accented phonotactic systems and the older the marion |
|---|
| 0:00:27 | the jfa on sites a five e |
|---|
| 0:00:30 | phil whether a myself often russell |
|---|
| 0:00:33 | we have a diversity of regional accents in the british isles and the papers about |
|---|
| 0:00:37 | trying to recognize fourteen of them |
|---|
| 0:00:39 | the that the of the work uses two different approaches one based on i-vectors and |
|---|
| 0:00:44 | the other based on the fusion of a set of fourteen different phonotactic systems so |
|---|
| 0:00:49 | the first is concerned with |
|---|
| 0:00:51 | acoustic differences between the different regional accents and the second between the is concerned with |
|---|
| 0:00:56 | differences in the sequences of phones that are produced in addition accents |
|---|
| 0:01:00 | unsurprisingly the best performance is obtained by using the two methods together |
|---|
| 0:01:05 | we have recognition results but also we have an attempt to visualise the i-vector space |
|---|
| 0:01:10 | where the actions wreck represented it is interesting to look at relationships between the topology |
|---|
| 0:01:15 | of that space and geographical and social relationships between the actual accents themselves thank you |
|---|
| 0:01:21 | again i apologise for not having a slot |
|---|