Case Study: Fortune 500 Tech Giant Company improves sentiment classification accuracy with Mu Sigma

Optimizing Sentiment Classification Using NLP and Advanced Analytics

Fortune 500 Tech Giant Company partnered with Mu Sigma to improve its existing sentiment classifier, which suffered from weak data cleaning and low sentiment accuracy. The company needed a more robust way to identify and prioritize issues in large volumes of customer feedback and reviews, but its current tool was not delivering reliable results.

Mu Sigma used its AoPS-based muOBI approach to revamp the sentiment classification workflow with RNN-based repunctuation, Stanza NLP for multi-target detection, a hybrid Stanza and deep learning model for aspect extraction, and a TASBA BERT model for target-level sentiment extraction. Mu Sigma’s solution improved overall accuracy from about 80% to 91%, delivered roughly 50% better text-to-sentence conversion, and improved aspect identification by about 30%, while also adding multi-target detection and target-level sentiment tagging.