This is a Python implementation of our work on customer care automation using a deep learning framework. We treat this problem as a question-answering problem, where the goal is to retrieve/generate an answer for a given question from customer.
We first learn the sentence embedding for questions and answers using Doc2Vec (an extension of word2vec), and then use a deep similarity network which gets a pair of question and answer and generates a similarity score.
This code uses Gensim package for Doc2Vec training, and Tensorflow library for Similarity Neural Network training.