Help in NLP Problem

Question

I am trying to solve an NLP problem and, since I am new to NLP I would like to ask for some insight.

I have a data-set with 300000 and 14 features. This data set is about customer complaints, and the data set includes the company the complaint was made, the date, the issue (and sub-issue), the product, and most importantly, the customer complaint narrative.

I am now supposed to build a classifier that provides the root cause of the problem, given an unclassified complaint narrative.

The thing is, there is an apparent root cause (I assume that what is in the issue feature) and the actual root cause, that may differ from the apparent root cause, and is contained inside the customer narrative.

I already tokenized,removed stopwords and used Lemmatisation on the customer narrative but now I am a bit lost as to what to do next.

Anybody can point me in the right direction?

Thank you!

EDIT: Added some text

Answer 1

Next steps should be :

1. Build a "Bag of Words" classifier as a baseline. Example : https://medium.freecodecamp.org/text-classification-and-prediction-using-bag-of-words-8aeb1396cded
2. Build a classifier with "Doc2Vec"

These should help with solving the problem. For example, "phone keeps dropping calls" might point to one set of root cause and bag of words should classify based on "call" and "drop" tokens.

Both of these (Bag of Words and Doc2Vec) are baseline models. Based on performance of these models, you will have to decide if it is good enough for the project. IF not, you will have to try more advanced models.

https://blog.francium.tech/content-based-text-classification-with-doc2vec-and-tensorflow-efd1dd4f02a8

https://towardsdatascience.com/multi-class-text-classification-with-doc2vec-logistic-regression-9da9947b43f4

https://github.com/ibrahimsharaf/doc2vec