The code I have shared in this blog enables us to capture real time audio conversation and do analysis of the conversation on the fly.
The list of analytics that it does are:
· Sentiment Score or Polarity Score
· Polarizing Words – The words that are contributing towards the dialogue’s sentiment score. We have divided the sentiments in 3 types – positive, negative and neutral.
· Keywords – The particular brand, company, person etc. who is being talked about here. The Proper Nouns after POS (Parts of Speech) analysis were marked as keywords.
· Context – The context of the conversation or the general subject area. The nouns (other than Proper Nouns) after POS analysis are marked as Context.
· Action Words – The action words associated with the conversation, i.e. the verbs.
For example : Let’s consider the following customer feedback –
"Disappointed that the Dell Outlook ticket was closed reopen the ticket."
Things worth noting the code analyzed are -
Here the polarity or sentiment score is -0.48.
Negative word identified: disappointed.
Keywords identified: Dell, Outlook
Context identified: Ticket
Action words identified: Reopen.
The script can be used at real time to connect to or alert appropriate teams who might be aware of the scenario and better suited to handle the issue. And if you are looking for text analytics it creates an immediate analytics without having to run a separate script later.