Skip to content

Voice-of-Customer NLP Pipeline

A VoC pipeline that mines topics, aspect sentiment, and trend shifts from reviews, chats, and NPS verbatims.

TransformersPythonFastAPINLP
Abstract teal and mint cover illustration for the Voice-of-Customer NLP project, showing sentiment waves and topic clusters flowing over a teal depth field.

Highlights

  • Hybrid topic modeling with LLM-labeled clusters and human review.
  • Aspect-based sentiment with product-taxonomy alignment.
  • Weekly trend digests with anomaly detection and drill-through.

Outcomes

  • Surfaced 5 previously-hidden churn drivers in the first month.
  • Cut manual VoC tagging effort from 20 to 2 analyst hours per week.

Stack

  • Transformers
  • Python
  • FastAPI
  • PostgreSQL
  • Airflow