Insiders & Debeka: Cognitive Classification with Deep Learning

Togeth­er with our cus­tomer Debe­ka, we report on the pro­cess­ing of image-based doc­u­ments in input man­age­ment with a new method at the spe­cial­ist con­fer­ence “AI and self-learn­ing sys­tems in insur­ance com­pa­nies” of the Ver­sicherungs­foren Leipzig. For the first time, image- and text-based clas­si­fiers were com­bined in an inte­grat­ed approach. The results are impres­sive and take the insur­ance group’s input man­age­ment to a new lev­el.Thank you for read­ing this post, don’t for­get to sub­scribe!

Under the title “Cog­ni­tive Clas­si­fi­ca­tion – Auto­mat­ed Image Recog­ni­tion with Deep Learn­ing Tech­nolo­gies”, we will present the joint project at the con­fer­ence. For many years, Debe­ka has been using Insid­ers Tech­nolo­gies’ solu­tions very suc­cess­ful­ly for divi­sion sep­a­ra­tion, clas­si­fi­ca­tion and pro­cess­ing of incom­ing doc­u­ments.

How­ev­er, the cor­rect recog­ni­tion of image-based incom­ing doc­u­ments such as ID cards, cer­tifi­cates or pho­tos of acci­dent dam­age or real estate proved to be dif­fi­cult or even impos­si­ble for the pure­ly text-based clas­si­fiers. Con­se­quent­ly, some of these doc­u­ment class­es had to be clas­si­fied com­plete­ly man­u­al­ly until now. This gave rise to the idea of extend­ing the clas­sic clas­si­fi­er to include image clas­si­fi­ca­tion using Deep Learn­ing. The new Cog­ni­tive Clas­si­fi­er from Insid­ers com­bines the two clas­si­fi­ca­tion meth­ods of text and image recog­ni­tion intel­li­gent­ly and seam­less­ly.

“Togeth­er with our part­ner Insid­ers Tech­nolo­gies, we have imple­ment­ed an AI solu­tion in a real use case that is absolute­ly in tune with the times.“

explains Marc Drew­ing, Group Head of Input Man­age­ment at Debe­ka Ver­sicherungsvere­in a.G.. “We can now use our proven process­es even more effec­tive­ly and fur­ther increase the lev­el of automa­tion.”

The Cog­ni­tive Clas­si­fi­er pro­vides an enor­mous automa­tion effect at Debe­ka: In the Home Loan and Sav­ings divi­sion, where most images accu­mu­late, the re-entry rate is now less than 16 per­cent.

“The Cog­ni­tive Clas­si­fi­er now pro­vides a uni­ver­sal solu­tion for the clas­si­fi­ca­tion of doc­u­ments and pho­tos that puts all exist­ing approach­es to shame,” adds Chris­t­ian Bokel­mann, autho­rized sig­na­to­ry and head of Key Account Man­age­ment at Insid­ers Tech­nolo­gies. “The project is already gen­er­at­ing a great deal of inter­est in the indus­try.”

The Ver­sicherungs­foren Leipzig GmbH face-to-face event will present reports and prac­ti­cal appli­ca­tion exam­ples from insur­ance com­pa­nies and beyond on the top­ics of AI, self-learn­ing sys­tems, machine learn­ing, auto­mat­ed deci­sion-mak­ing and Explain­able AI on May 17 and 18, 2022 in Leipzig. It is aimed at spe­cial­ists and man­agers from the cen­tral and decen­tral­ized areas of insur­ance com­pa­nies, par­tic­u­lar­ly in the depart­ments of IT, process man­age­ment, oper­a­tional orga­ni­za­tion, data man­age­ment, sales and mar­ket­ing, and claims.