This makes us unique
Our speech recognition has a number of functionalities that are unique in the industry. Read here what you can only get from us.
By replacing all personally identifiable information in the speech-recognized text, with labels indicating what information has been removed, conversations can be anonymized in real time. Thus, no audio recording of the conversation is made and the anonymised text can be saved without being in conflict with GDPR legislation.
Real-time or batch anonymization of audio
Personally identifiable information, identified by text anonymization, can be replaced with a tone. In real-time recognition, this can be done in memory so that the un-anonymized conversation is not recorded on any media. It is thus possible to save audio recordings for as long as you want, without being subject to GDPR considerations.
All significant words, in speech-recognized text, are weighted with a value from –5 for the most negative to +5 for the most positive.
The sentiment values for all the significant words are summarized and divided by the number of significant words. This achieves an overall sentiment value for the conversation.
The AudioStreamer application sends audio to the speech recognizer in the best quality, directly from the Agent's PC - without telephony integration. Audiostreamer is a Windows app that streams audio from both agent and customer directly from the sound card to Mirsk speech recognizer, or to a Secure FTP (SFTP) server if there is no need for real-time recognition. Thus, there is no need for telephony integration.
Delivery of transcribed text in structured database
Transcribed text can either be provided as a text file for each conversation or audio file, or in a structured MS SQL database format. In addition to the text, the database will contain timestamps for start and stop of each word, sentiment score per word (if this option is selected), speech ID at 2 track sound m.m.