NLP Data Pipeline

Customer Feedback
Intelligence Analyzer

Processing 10,000+ unstructured reviews using VADER sentiment analysis and BERTopic clustering on TF-IDF embeddings to surface actionable business pain points.

Python 3.9+ BERTopic VADER Sentiment Flask API

Performance Metrics

10,000+
Reviews Processed
9,100/s
Pipeline Throughput
34.8s
Total Runtime
69.3%
VADER/Star Agreement
44
Discovered Clusters
10
Actionable Insights

Interactive Dashboard

A scalable Flask-based visualization system highlighting sentiment trends, topic heatmaps, and category breakdowns.

Positive Sentiment 60.9%
Negative Sentiment 31.8%
Neutral Sentiment 7.3%

Sentiment Distribution

Sentiment by Category

Top Discovered Clusters (BERTopic)

Pipeline Architecture

1. Data Ingestion Layer
Raw CSV
Cleaner
Preprocessed DataFrame
2. NLP Processing Pipeline

VADER Sentiment

  • Compound
  • Positive
  • Negative

BERTopic

UMAP HDBSCAN c-TF-IDF
3. Analysis & Visualization
Sentiment Dashboard
Topic Heatmaps
Pain Point Extraction