ExB Labs GmbH with its subsidiaries is a German research and technology company for Artificial Intelligence (AI) and develops core technologies in the areas of NLP and Machine / Deep Learning. The company specializes in the production of self-learning software platforms for the analysis of large quantities of unstructured and partially structured text and image data. ExB's software solutions are not only web and cloud based, but also embedded and applicable on mobile devices.
The ExB AI platform can configure, train and deploy processing modules. While industry-unspecific, it has been applied in the insurance, health and mobility sectors. The product, called Cognitive Workbench (CWB) is self-learning software for the processing and analysis of unstructured text and partially structured data such as tables and invoices as well as image / document data.
The CWB outputs a docker to easily deploy it for processing large data volumes. Consequently the product is cloud and on-premise ready, which makes it well suited for industries that handle sensitive personal data, such as insurants, patients and drivers. ExB cooperates with companies as Allianz and Talanx in the insurance realm, Bosch and Audi in automotive or Springer-Medical in health.
ExB, meaning External Brain, is in currently in the process of optimizing the product-market fit for their insurance offering. 15 German Insurers are testing, implementing or using the platform.
ExB’s CWB helps increasing customer satisfaction while saving costs
ExB applies AI to solve an age-old problem: Automated processing of incoming communication. The handling of incoming documents is one of the highest cost drivers in claims processing and policy administration. Traditionally, this process is managed by combining specialized system for OCR, rule-based extraction and classification mechanisms and many manual processing and correction steps in between.
The ExB Cognitive Workbench uses latest NLP (Natural Language Processing) techniques incorporating the best fitting machine learning paradigms to holistically process incoming communication, thereby reducing the usual error stacking that occurs when separate systems for OCR, classification and extraction are used. The system is self-learning and thus able to handle cases, which might not have been foreseen and trained. Initial trials have shown substantial quality improvement as compared to the traditional approach of selecting multiple software suppliers covering the processing chain.
How it works
Usually, the handling and processing of incoming mail at insurance companies is managed by a combination of various in-sourced, monolithic sub-systems. These include special tools for optical character recognition (OCR) and rule-based extraction and classification mechanisms. In between there are various manual processing and correction steps. They do not have self-learning algorithms that improve results and recognition rates during a training process. With the help of machine learning based systems like the Cognitive Workbench, the essential information from the correspondence can be recognized, extracted, correctly assigned and stored and this also in mutual interactions to each other.
The system is actually understanding a document/incoming mail to a full extend and can process it automatically. The complete automation of the workflow enables insurance companies to save resources, effort, time and costs and to provide superior customer service.
The Infographic illustrates the process flow at the customer for closing the data loop.
The screenshot shows the configuration of data source, analytics and output with simple mouse clicks, no programming skills required.
Why we selected ExB for DIA Amsterdam
The Cognitive Workbench (CWB) uses latest NLP techniques incorporating the best fitting machine learning paradigms to holistically process incoming communication. The system is fully scalable, self-learning and thus able to handle cases, which might not have been foreseen and trained.
At DIA Amsterdam, Dr. Ramin Assadollahi, CEO and founder of ExB will present an example of the power of the Cognitive Workbench handling a specific case of client communication within an everyday claim-example.
Who is ExB ?
ExB was founded in 2000 by the computer linguist and clinical psychologist Dr. Ramin Assadollahi. The ExB Group currently employs more than 70 people at its locations in Munich (headquarters) and Leipzig. Late last year they closed a funding round of € 9.5 million. ExB has won many prices like: 2017 (text and scanned documents): Allianz Hackathon “OCR meets Machine Learning” and the ISBI / ISIC 2016 (medical images): First place for „automated segmentation of melanoma from dermoscopic images“ and many more.
The first product preDICT was a next-word prediction based text input for smart- and feature-phones that shipped in 70+ languages. preDict was sold in an asset deal to Nokia, the technology is in use in more than 100 Mio phones today.
”ExB means External Brain, isn’t that enough of a vision?” Dr. Ramin Assadollahi, CEO and founder of ExB