The problem

I have a problem where I have text data that has been transcribed from a conversation. These conversations have been marked as a pass or fail in terms of compliance by a person, ie they have listened to the call and marked the call as passed compliance or not passed. I am trying to think of a way to use text data in regard to analytics. The conversations are just text data from a staff member to a customer, so I was thinking of trying to turn this into a learning problem to try and score conversations in terms of the risk. To explain further, score the conversation as the probability that the call passed or failed based on the text data. My Thoughts

I was thinking of trying to create a model that can score new conversations based on previously labeled data as we have calls that have already been scored as a pass or fail. I am just not sure if this solution is possible given the current technology stack.

My Questions-

  • I assume the main question is, is there a correlation between the conversation text and a call not passing compliance. How would I be able to calculate this correlation?
  • Would this be possible, what type of data pre-processing would need to be done.
  • I think as its text data the data pre-processing is the most important part. Information on the pre-processing steps would be much appreciated.
  • Is there an example online that can be used as a reference?
  • Would there be any other analytics that would be more useful? ie. a different type of solution.
  • what size dataset would be suitable, just to get started.

If possible can you please add links to articles, tutorials, books, etc.?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.