A Proof-of-Concept Analytics Platform for Voice of Customer Feedback
VoC Pulse is an AI-driven analytics proof-of-concept (POC) designed to analyze unstructured customer feedback quickly and effectively. Originally developed during a 48-hour sprint, the project provides capabilities to:
- Process customer feedback to extract insights on Sentiment (e.g., Positive, Negative, Mixed).
- Identify Topics (e.g., "Buffet", "Service").
- Apply Vector Embeddings (semantic search) for deeper analysis.
This POC leverages key tools and libraries for rapid data visualization and analysis, presenting insights interactively.
Below is a dashboard showcasing some of the outputs generated by the system:
- Interactive Dashboards: Visualization of customer sentiment trends using Plotly for interactive trend charts and Matplotlib for distributions.
- Semantic Pie Charts: Charts summarizing customer sentiment breakdowns by selectable criteria.
- Keyword Generators: Generate dynamic word clouds reflecting common keywords in feedback (e.g., "excellent", "delay").
- Prototype Retrieval-Augmented Pipeline: A manually implemented RAG (Retrieval-Augmented Generation) system combining OpenAI API and ChromaDB for querying relevant feedback data.
This POC is in its early stages, and functionalities such as a complete API backend and agentic capabilities are yet to be developed. The current version is functional as a local app for demonstration of core concepts. Future updates will include:
- Robust API endpoints (using FastAPI).
- Autonomous agents for orchestrating insights generation.
- A full-fledged dashboard for interactive stakeholder discussions.
For a demo or further development, refer to the roadmap in development notes.
This project is licensed under the MIT License. See the LICENSE file for details.