Speech software to test refugees' claims of origin

German authorities plan to use automated software to analyze the dialects of asylum seekers to verify their claimed place of origin, national broadsheet "Die Welt" reported on Friday.

It reported the Federal Office for Migration and Refugees (BAMF) planned to start trials of the speech analysis software within two weeks and a full roll out in 2018.

"The idea is to record a separate speech sample from asylum seekers and to carry out an automatic dialect analysis," Julian Detzel, of BAMF's Global Strategy digitization and IT program management told "Die Welt."

The software, which was based on the same voice authentication technology used by banks and insurance companies, would help migration officers review the applicants' sources of origin as one of several "indicators," the paper reported.

Since 1998 Germany has used speech analysis to assess claims of origin. In cases of uncertainty, recorded clips of conversation with the applicant are sent to a linguistic expert who can listen for dialectic variations such as differing names for food.

'Humans and machines can easily be wrong'

Professor Monika Schmid, a linguistics expert who studies "language attrition", particularly among migrants, told Deutsche Welle such analyses can be fraught.

"Identifying the region of origin for anyone based on their speech is an extremely complex task," the University of Essex professor said. 

"We have argued that in order to do so reliably, an analyst must have a solid background in linguistic analysis and be able to take into account a wide range of factors. For example, people will adapt the way they speak to the speech patterns of their interlocutors. 

"I don't see how automated software can distinguish whether a person uses a certain word or pronounces it in a particular way because this is part of their own repertoire or because they were primed to do so by the interviewer or interpreter."

Her team recently published a study where native German speakers listened to speech samples and decided whether the speaker was a native speaker. All the samples came from native speakers who had lived abroad for at least five years, but the test group identified many of them as being foreign speakers.

"Crucially, in this context, we also did an acoustic analysis of the pronunciation of each speaker (which, I imagine, must be similar to what automated analysis does). Again, for many speakers the analysis identified non-native pronunciation,"  Schmid wrote.

"So, both humans and machines can easily be wrong, but humans are probably better at realizing this."

BAMF had 45 experts for a total of 80 languages "German daily Süddeutsche Zeitung" reported in 2016. It reported that one third of analyses confirmed the claimed country of origin. In 2015 411 such analyses took place.

aw/rt