Germany to use automated speech analysis for migration applicants

By AT editor - 17 March 2017 at 6:10 pm
Germany to use automated speech analysis for migration applicants

Within two weeks, German authorities plan to use automated software to analyze speech patterns of  asylum seekers to verify their claimed place of origin, according to a Deutsche Welle report.

The software is based on similar authentication programs used in the finance and insurance industries to assess speech patterns when confirming identity.

Deutsche Welle, citing the German daily “Die Welt,” said the Federal Office for Migration and Refugees (BAMF) will begin using the automated dialect analysis software with full implementation planned for 2018. The results will be used by migration officials as one factor when considering migrant applications.

Germany has used speech analysis for nearly 20 years on a more limited basis. If uncertainty about an applicant’s stated country of origin arose, recordings of conversation with the migrant would be sent to a linguistics expert for review.

The assessment would include language variations and accents – for example, noting that a French or Arabic speaker demonstrated key dialect markers – or by the specific use of common words for food, clothing or other clues rooted in culture.

Critics warn of a number of potential problems with both automated software or human linguistic analysis. For example, a subject will commonly alter responses in ways expected by the interviewer, or on the basis of cues they’ve understood about the process.

Both automated systems and humans can be wrong in their assessment, but as linguistics expert Monika Schmid told Deutsche Welle, the humans are “probably better at realizing this.”

In 2016, just a third of the linguistic analyses conducted by BAMF confirmed the stated country of origin.

1 Comment so far

Jump into a conversation

  1. Pingback: March 2017 – Migration Mapping

Leave a Reply


This site uses Akismet to reduce spam. Learn how your comment data is processed.