Germany plans to use speech analysis technology to help determine asylum seekers’ countries of origin, according to a report from Die Welt. The Federal Office for Migration and Refugees (BAMF) will begin testing the software within the next two weeks, the paper reports, with an eye toward deploying it more widely in 2018.
The hope is that the software will be able to analyze and identify the dialects of people seeking asylum in Germany, based on recorded speech samples. That information could then be used as one of several “indicators” that migration officers consider when reviewing asylum applications. The technology is based on voice authentication software used by banks and insurance companies, and will be modified to analyze dialects, according to Die Welt.
BAMF has previously estimated that around 60 percent of people who sought asylum in 2016 did not have identification papers when they arrived in Germany. In February, the German Interior Ministry presented a draft law that would allow migration officers to seize smartphone and laptop data to help determine asylum seekers’ identities, in a move that raised concerns among some privacy advocates.
“Creating a perfect dataset is virtually impossible.”
Germany has been using linguistic experts to analyze dialects and determine countries of origin since 1998, according to Deutsche Welle. But experts have expressed doubts over whether computers would be able to reliably perform such analysis.
Monika Schmid, a linguistics professor at the University of Essex, tells Deutsche Welle that speech analysts “must have a solid background in linguistic analysis and be able to take into account a wide range of factors” when determining someone’s country of origin, including changes in the ways they speak around different people.
“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," Schmid told Deutsche Welle.
Dirk Hovy, a computer scientist at the University of Copenhagen, tells Die Welt that BAMF’s system would need to incorporate speech data that is demographically representative of asylum seekers, which would involve the creation of a broad database. “Creating a perfect dataset is virtually impossible,” Hovy said, “because language is constantly changing.”