Skip to content
Topic Classification and Sentiment Analysis for Generating Social Sentiment Score to Aid Credit Rating

Objective:

How to analyse comments from various sources on companies (and their competitors) to track sentiments associated with them. These will serve as inputs to the credit scoring engine to take into account social sentiments as a soft input for ratings.

 

Key Challenges:

  • Presence of various sources of unstructured data:
    • Customer Reviews
    • Company Surveys
    • Technology Blogs
    • Social Media Posts
  • Lack of a comprehensive scoring method, combining all the above sources of information
 

Approach:

  • Corpus of vectors (words) made from raw and unstructured data
  • Pre-processing engine for white-space removal, punctuation-removal, stop-words removal, etc.
  • Term document matrix creation
  • Text Classification and NLP Algorithms are used to categorize each comment into classes
  • Sentiment Classifier engine using an ensemble of algorithms to arrive at a weighted sentiment score with optimal weights to reduce errors
 

Benefits:

  • Provided inputs for a comprehensive credit scoring methodology that helps take into account social sentiments associated with the brand as a risk/saliency factor