Multilingual customer sentiment engine
NUT Botanicals · Consumer Feedback / Marketing
Tracked over the deployment window versus the prior baseline.
- Client
- NUT Botanicals
- Industry
- Consumer Feedback / Marketing
- Scope
- Multilingual customer-sentiment platform across channels.
Inside the project
Challenge
NUT struggled to make sense of scattered customer feedback across platforms. The goal was to understand sentiment and drive product strategy.
Solution
A multilingual sentiment-classification engine that aggregates feedback from social media, e-commerce platforms, and surveys, then classifies it with machine-learning models (Naive Bayes, SVM, BERT) to support strategic decisions.
Our approach
We consolidate feedback from social media, e-commerce and surveys, then classify sentiment and topics across Arabic and English — Naive Bayes and SVM give explainable baselines while BERT captures richer context, sarcasm and product terminology. It runs an operating loop — ingest, anonymize, classify, aggregate, alert, assign, resolve and track each topic over time — surfacing topic trends, complaint categories, product-issue heatmaps and urgent-feedback queues, not just a positive/negative score. Language-quality review, bias testing, anonymization and escalation handle ambiguous or high-impact feedback, so emerging issues are caught early.