Speech Analytics Starts with Call Recording
This guest post was authored by our friends at OrecX.
Speech analytics technology requires quality audio going in for quality transcription and analytics to come out. This all starts with the call recording itself. The audio must be crystal clear or the transcription will be off and the resulting intelligence will be flawed.
The clarity of the recorded call comes down to several factors, including the audio acquisition itself. Most call recording solutions capture telephone calls in mono format. This means both parties on the phone are recorded on the same channel. This can cause problems when it comes to transcription. The system cannot always clearly distinguish between the two parties, especially if both individuals are talking simultaneously. Suppose a potentially costly dispute arises (perhaps over authorization to charge a credit card). If the transcriber technology cannot discern whether or not the customer did in fact provide approval, the company could be on the hook for the charge.
Stereo recording, on the other hand, captures the two parties on the telephone on separate channels. This leads to much higher fidelity audio, which makes it easier for the automatic call transcriber to accurately distinguish who said what in the transcription. Stereo (aka dual channel recording) also greatly minimizes background noise, such as a dog barking, kids playing or coworkers chatting too loudly. With dual channel and the minimal external noise, the automatic transcriber can very accurately transcribe the call, and the speech analytics engine can then precisely spot keywords to help your contact center with quality assurance, compliance, dispute resolution and order verification.
There are other factors at play as well when it comes to ensuring the highest quality transcription and speech analytics analysis, including:
- High quality codec
- Access to multiple codecs
- Ability to add meta data to a call
- Meta data filtering
- Portability of recorded call and data
- Open API for easy interoperability
If you are considering a call recording provider and have an interest in the tremendous value speech analytics can provide, then you need to know what specific questions to ask the call recording vendor(s) you are considering.
Questions to Ask a Recording Vendor
- Do you record calls in mono or stereo?
- If stereo, do you keep the channels separate upon playback? [This is very important]
- Is your integration API-driven for easy connectivity to both post-call and real-time analytics
- Is the audio in your recordings compressed or uncompressed?
For the purposes of feeding the speech analytics solution, the call recorder you choose should be able to run in parallel to your existing recording solution and simply be used as the engine to feed the analytics system.