A LSTM based Tool for Consumer Complaint Classification

The work focuses on the issues in handling consumer complaints and queries in online platforms. In this digital era, the majority of complaints are posted in blogs, forums and social networking pages of companies. Users may not use official company websites or platforms to submit complaints. Thousands of queries and complaints are posted in digital platforms on a daily basis regarding different issues. It's a tedious process for a human to go through a database of such queries and manually tag each and every query to it's correct class or category. The consumer service representative who is handling that department can start working on the issue only after the tagging process. The manual intervention in this scenario is very high. A Long Short Term Memory (LSTM) based tool is proposed to automatically tag consumer queries to its corresponding class. Also, it identifies outliers or queries which are not related to the company issues or matters. The minimal manual intervention will reduce the response time. The model showed considerable accuracy when evaluated with validation data.