A Jointventure of
 
idw news
A | A | A
 

Efficient and Effective: Speech Recognition and Text Analysis in One

At CallCenterWorld in Berlin, the European Media Laboratory will be presenting a combination of robust speech recognition and sophisticated text analysis tools - "Domain adaptation" enables enterprises to tune speech recognition performance for their own applications and thus to respond quicker

Berlin/Heidelberg. The Heidelberg IT enterprise EML European Media Laboratory will be represented at this year's CallCenterWorld in Berlin (28 February to 1 March 2012) with a stand of its own (Hall 5, Stand A8). EML offers enterprises speech technology solutions for their customer contacts over the phone. The "EML Transcription Platform" provides automatic transformation of incoming phone calls into written text, analysis of the calls, and the tuning of the speech recognition performance to customer-specific application domains.

At CallCenterWorld, EML's speech technology experts will be demonstrating a novelty: "domain adaptation." It enables users to adapt speech recognition to their own application domains at short notice and without assistance. For instance, an enterprise launching a new product can quickly and effectively insert product descriptions or new keywords into the language model. Another unique selling point that EML has to offer is that speech recognition and domain adaptation capabilities will be available to enterprises on their own premises. "That safeguards data protection issues," says Dr. Siegfried Kunzmann, head of research and development.

Incoming phone calls are automatically transformed into written text and classified in accordance with call-center-specific product or problem categories, patterns or conversational features. In addition, the calls are scanned for new terms and linked up with one another accordingly. The automatic transcription now boasts "advanced keyword spotting," an extension of the search function for particular keywords. In conventional systems, every new keyword search involves reprocessing and rescanning all incoming calls that have been recorded. "Advanced Keyword Spotting finds the keywords in recognized texts as quickly as a search engine," says Dr. Kunzmann. "And the new phonetic search function means we can find keywords not yet registered in the speech recognition system, for example unusual proper names."

Another asset of text analysis is to establish effectively how many callers have referred to particular product names - interesting information for a company's strategy and marketing department. Phone calls are authentic customer statements that can be drawn upon to modify products, services, and activities faster and more directly. "Our combination of robust speech recognition and sophisticated text analysis tools helps call-center workers to make communication more efficient and customer-friendly," says Dr. Kunzmann. "And it provides a tool for strategic marketing."

28 Feb. to 1 March 2012 at CallCenterWorld, Berlin, Estrel Convention Center: Hall 5, Stand A8

About EML EML European Media Laboratory GmbH was established by Klaus Tschira, a co-founder of SAP. Research and development at EML centers on automatic speech recognition, speech processing, and man-machine interaction. Specialties are speech-to-text applications, as used in the telephone sector and in call centers (speech analytics, voice mail), and the transcription of media data.

For further information, please contact

Dr. Peter Saueressig Public Relations European Media Laboratory GmbH Schloss-Wolfsbrunnenweg 35 D-69118 Heidelberg phone: +49 6221 533245 saueressig@eml.org

idw :: 08.02.2012