20.02.2018

Real-time speech analytics: benefits for customer contact

EML European Media Laboratory is presenting the advantages of automatic transcription and machine learning for call centers at the international trade show CCW in Berlin. The language technology company will also demonstrate its expertise in language assistance systems which can support the work of call center agents.

Offers advantages for the call center: real-time transcription of calls. Photo: Yuri Arcurs

Berlin/Heidelberg. Once again this year the Heidelberg-based IT company EML European Media Laboratory GmbH will have its own stand (Hall 2, Stand C19) at the telecommunications trade show CCW. The show will run from February 26 to March 1 in Berlin and is expected to attract more than 7,000 visitors. The Heidelberg speech recognition experts will present applications for telephone customer contact that transcribe incoming calls into text in real time. The recognition results can then be further processed, quickly and easily. They will also demonstrate a language assistance system that supports state-of-the-art privacy protection and works on-premise, i.e. without a link to an external service provider.

State-of-the-art technologies for speech recognition

The EML Transcription Server immediately converts incoming calls into machine-searchable text, automatically and speaker independent. "We use the most up-to-date machine learning techniques research has produced," explains EML Managing Director Professor Andreas Reuter. "With neural networks and deep learning, we achieve the highest recognition rates." The EML Speech Mining Server then classifies the results according to pre-defined categories, the frequency of certain expressions, and keywords; it identifies new terms that occur frequently and combines the results. Real-time speech recognition also makes it possible to simply and efficiently verify and monitor compliance with conversation guidelines at the call center. The system works in several languages: in addition to German and English, EML offers speech analytics among others in Arabic and Chinese.

Automatic key-word search

The automatic transcription of calls into text facilitates the efficient, extended search for certain keywords. With every new keyword search, conventional systems have to re-process and check all calls. Advanced Keyword Spotting, on the other hand, finds keywords in the converted text as quickly as a search engine. "We also support traditional Keyword Spotting (KWS)," explains EML Development Manager Markus Klehr, "but from our point of view, Advanced Keyword Spotting does much more: it lets you connect the keywords with one another."

"Your Customer´s Voice": customized marketing tool

Call center operators can adapt speech recognition to their own application domains. When a new product is launched, for instance, the product description and new keywords can be added to the language model. Immediate issues or emerging trends are thus quickly recognized.
Employing speech recognition also allows the call center to elucidate exactly how many callers have cited a certain product name – interesting information for a company's strategy and marketing departments: after all, the calls are authentic customer statements which can be used to adapt products, services and activities more quickly and directly. "This helps employees in call centers to be more efficient and customer friendly in their communication with callers," says Markus Klehr.

Using language, protecting data

At CCW, EML is also presenting a language assistance system which can be used both within your own four walls and for corporate purposes. What is special about it is that it works without a permanent data connection to the "outside world" – the speech recognition is installed in the device itself. Users can call up the entire spectrum of voice-enabled communication undisturbed: search for news, weather and traffic data online, control heating, electric blinds and lights, call up music and other services. Their data, however, remain on-premise, in their own homes.
Of course, we also apply the same high data protection standards to domain adaptation in call centers," says Markus Klehr. "Recognition and model adaptation are conducted on-premise, which means call centers can comply with our strict European privacy protection law."