Can we make sense of what is said?
When the message is not enough
we help extracting meaningful spoken features
that describe emotional states
What did she say?
A lot of customer interactions take place today through audio channels, either human assisted or automated. In order to properly process the messages issued by the customer a process is needed to convert speech to written text.
Many times this process needs to cater for local particularities and context in order to increase the accuracy of the speech-to-text conversion.
How did he say it?
But the what is not enough. There same string of words can be percieved as kind, fearsome, false etc depending in features that can be extracted from the audio.
Changes in pitch, silences and various emotional accents can add meaning to the text-only message. This additional information can be in turned exploited to tailor the response and improve the customer experience
Historical audio data is needed to produce a high performing speech analysis system.
If you were able to get automated insight from oral communications, either in real time or in a batch post-processing step, don't you think you could take some advantage out of it?
A seasoned collection agent is able to ascertain if a promise to pay is believable or dubious. What iif this assessment could be automated?
If you were able to differentiate true interest to buy from information requests of no consequences, wouldn't you focus your efforts and investment in that low hanging fruit?
Your speech analytics system can help you in these directions