So, I have users complaint data for an automobile firm. I want to map each complaint to a specific topic based on the context. For example : Complaint Description : As per telephonic conversation with customer, customer said that on 26th of Oct. 2009 and from the very first day he is facing starting issue in his vehicle, regarding this issue he visited service center, there they replaced the starter motor but again he is facing the same issue in his vehicle, customer is very disappointed, he wants replacement of vehicle, he wants to raise the complaint, hence complaint raised. "
Output : Topic : Engine/vehicle start issue
In topic , I want to capture only the issue he is facing.
Is there any NLP/ topic modelling approach that can cater these kind of use cases ?? Any useful links would be also appreciated.
I tried word cloud on the corpus I had and based on that, selected some frequent words/phrases to assign topics to each complaint. So, if a complaint consist of a word 'gear stuck', then will assign it to 'gear stuck issue'. {'engine start', 'vehicle stuck', 'engine stopped'} --> 'engine start problem' {'oil leak', 'coolant', 'oil flowing'} --> 'oil leakage'
But this process is not a standard one. I want a better approach to do this.