Munich – 26.11.2014
ExB: Cutting edge innovation in natural language processing
First and second prizes go to ExB teams
PTPT technology platform proves its lead in Named Entity Recognition
Munich-based technology development company ExB (www.exb.de) has received an impressive confirmation of the effectiveness of its innovative language technology. As a participant in the GermEval Challenge – an international competition for Named Entity Recognition technology (NER) – the ExB team took both first and second place with two different submissions.
The goal of the competition was the most accurate identification and recognition of names and the categories they belong to (like people, organizations, and places) in a large collection of German texts. A total of 11 teams participated in the competition with 33 submissions. The two solutions ExB submitted to GermEval rely on two different Named Entity Recognition algorithms: The first place winning solution uses a Conditional Random Fields (CRF) approach, while the second is based on maximum entropy methods.
The Big Picture for Big Data: PTPT makes possible the analysis of unstructured data
Computational linguistics and automated name recognition are key technologies for data processing that support innovations in computer-aided translation, data mining, natural language processing and biomedical science. Big data analysis largely relies on the kind of known, structured information stored in databases. The technology developed by ExB works with unstructured data like written texts, converting it into a structured form that is suitable for big data analysis techniques.
Natural language processing technologies are evaluated not only for their processing speed and ability to handle large quantities of data, but also for the quality and usefulness of their results. Furthermore, processing German language texts is a complex technological challenge in itself. ExB’s PTPT platform and its intensive development work in natural language processing is the basis for the unmatched accuracy of its technology.
PTPT: Universal platform for linguistic data extraction
The PTPT platform underlies all of ExB’s language technology. PTPT is an application for language and information processing and a universal platform for implementing rich data management solutions. It provides comprehensive solutions that go far beyond name recognition and natural language data extraction, including example-based image recognition, extracting text from images, and location-based data processing.
Dr. Ramin Assadollahi, founder and CEO of ExB, sees this first-class competitive result as strong evidence of how powerful PTPT is: “ExB sees itself as incubator for innovation in data extraction and processing. Our developments are at a very early stage in their commercial application, so it is extremely important in this initial phase to get scientific and objective confirmation of their excellence and competitiveness.”
The GermEval 2014 competition took place at the beginning of October as a part of Konvens 2014, held at the University of Hildesheim. This event is as an important meeting place for German computational linguists, bringing together 130 participants from over 60 institutions. Taking part in scientific competitions is an important activity for ExB in order to demonstrate the strength of the PTPT platform and its position in the language technology industry. These competitions display the diversity of PTPT’s solutions for language and data processing.