In Real Time: Automatic call transcription in call centers

The Heidelberg-based IT company EML European Media Laboratory GmbH will be presenting its products at this year’s CCW trade fair (hall 2, C19). The fair is taking place in Berlin from February 23rd to 25th. More than 7000 visitors are expected.

At the CCW, the Heidelberg speech recognition experts will show applications for telephone customer contact that allow for real-time transcription of incoming customer calls into text. The recognized text can be further processed fast and easily.

Newest technologies for speech recognition

The “EML Transcription Server” automatically converts incoming phone calls, speaker-independent and in real-time, into machine-searchable text. “We apply neural networks and deep learning techniques,” explains R & D manager Dr. Siegfried Kunzmann. “Thus we achieve excellent recognition rates.” In a further step, the “EML Speech Mining Server” classifies the recognition results according to predefined categories, calculates the frequency of certain expressions and keywords and finally links the results with each other. Real-time speech recognition facilitates the monitoring and efficient verification of conversation guidelines of call centers.

Automatic transcription is the key requisite for efficient, extended search for particular keywords. In conventional systems every new keyword search involves the reprocessing of all calls, that is of all recorded audio data. EML's “Advanced Keyword Spotting“, however, finds keywords in the recognized text as quickly as a search engine. The keywords can also be linked with each other. In addition, the phonetic search function will find words that are not yet included in the speech recognition vocabulary – unusual names, for example.

Individual customization and privacy

Call center operators can furthermore adapt the speech recognition to their own application domains. With the launch of a new product, the adaptation process allows to easily add product descriptions or new keywords to the language model. This makes it possible to quickly recognize current issues or emerging trends. Another advantage of this customization is that “speech recognition and model adaption take place on the companies’ own premises which guarantees data protection,” says Dr. Siegfried Kunzmann.

Voicemail: Reading is faster than listening

EML’s speech technologists also demonstrate the “EML Voicemail2Text Service“ and the corresponding Android app for automatic transcription of voicemails. This service is free of charge for private users. With the “EML Voicemail2Text Service” owners of a VoIP router, such as Fritz!Box, can receive an automatic transcription of incoming voicemails on their smartphones without additional installations on their router. They simply sign up at https://vmt.eml.org/ or use the „EML Voicemail App“ that can be downloaded for free at the Google Play Store.

23.02.-25.02.2016: The EML at the CCW, Berlin, Estrel Convention Center: Hall2, stand C19.

EML European Media Laboratory GmbH:
EML European Media Laboratory GmbH was established as a private IT enterprise by Klaus Tschira, a co-founder of SAP. EML develops software and technologies in automatic speech processing. The main focus is on the automatic conversion of speech into text (transcription) which is used for telephone (speech analytics, voicemail) and mobile applications (voice texting, voice search).
>www.eml.org>

Follow EML:
Twitter: @EMLvoice
Facebook: /EMLEuropeanMediaLaboratory
Youtube: /EMLVoiceMessaging
Google+: /+EMLVoicemailService

For further information please contact:
Dr. Peter Saueressig
Head of Communications
EML European Media Laboratory GmbH
Schloss-Wolfsbrunnenweg 35
69118 Heidelberg
Tel.: +49-6221-533245
Email: saueressig@eml.org

http://www.eml.org/english/media/presse.php?we_objectID=192 EML press release
http://www.ccw.eu/en.html CCW website
https://play.google.com/store/apps/details?id=org.eml.app.voicemail EML Voicemail App