4

I am looking for Consumer complaint/review description data which is not generated from automatic logs but manually entered by the end-user/customer-support.

The objective is to run the Stanford NLP-NER tool on the user description for determining the Location, Organization, Faulty Device, etc related to the complaint. I have been struggling for 2-3 days to explore and have found some Telco data, but all of the Data is STRUCTURED. See my answer on this question.

The key elements of the data I need are -

  1. UNSTRUCTURED description by the end-user or by customer-support.
  2. Data can be in the form of either of 'complete sentences' or phrases(say comma separated/semi-colon separated data). A hypothetical example (just for the sake of clarity and not restricted to the following format) of tow of the concerned columns among the rest of the user/caller info -

Fault Description - "I am calling from New York; staying at Hotel Marriot Marquis; Please help asap - my phone not working."

Fault Resolution - "System connection established in New York; Hotel Marriot Marquis; Issue resolved by on-site visit - problem in region multiplexer..bla bla."

I believe this problem would make for an interesting application in the Telecom domain.

Community
  • 1
user1412066
  • 81
  • 4

1 Answers1

0

I'm also looking for similar. Did you check reddit forums? Reddit has a decent API for scraping. Also many telcos have their own customer complaint forums.

Also structured, but I came across the fact that MeaningCloud provides some pre-trained categorization models for Telco vertical https://www.meaningcloud.com/developer/documentation/predefined/deep-categorization-models/voc-telco

Let me know if you found anything else!

dcsan
  • 101
  • 1