Sentiment Analysis for Moderated Usability Tests (Audio) ‐ Basma Elhoseny
GSoC'24 — Sentiment Analysis for Moderated Usability Tests (Audio)This project introduces audio-based sentiment analysis for moderated usability tests within the RUXAILAB ecosystem. The goal is to enrich usability evaluation by combining speech transcription and sentiment classification, allowing moderators to better understand users’ emotional responses during test sessions.
🔗 Data Extraction for Sentiment Analysis from Usability Tests
-
Basma Elhoseny
- Role: GSoC Contributor – Student Developer
- GitHub Profile
The project focuses on extending RUXAILAB with Sentiment Analysis for Moderated Usability Tests, specifically targeting audio-based evaluations.
It enables moderators and researchers to:
- Analyze emotional sentiment (Positive, Neutral, Negative) from user speech.
- Associate sentiment results with specific audio regions.
- Combine transcriptions and sentiment insights to better understand user experience during usability tests.
The solution integrates directly with the Sentiment Analysis API, developed as part of GSoC 2024.
The solution is composed of two tightly integrated parts:
Repository: ruxailab/sentiment-analysis-api
Stack: Python, Flask, Whisper, RoBERTa, PyTorch, Docker
- Audio extraction from video and audio files.
- Speech transcription using Whisper.
- Sentiment classification using RoBERTa.
- Timestamp-based sentiment analysis per audio segment.
- REST API with full documentation and test coverage.
- Docker and Docker Compose support.
Repository: ruxailab/RUXAILAB
Stack: Vue.js, Vuetify, JavaScript, Firebase, WaveSurfer.js
- Audio Sentiment Tab integrated into moderated test answers.
- Interactive audio waveform visualization with region-based sentiment analysis.
- Playback controls (speed and volume).
- Display of transcriptions mapped to sentiment per audio region.
- Improved Firestore querying with multi-condition support.
- 🎙️ Implementation of Sentiment Analysis for Moderated Usability Tests (Audio).
- 🧩 Integration with the Sentiment Analysis API.
- 🗂️ Creation of new models, controllers, views, and UI components.
- 🔎 Improved Firestore querying via reusable multi-condition queries.
- 🎨 Enhanced UX for moderators through intuitive audio visualization.
- Frontend: JavaScript, Vue.js, Vuetify, Firebase, WaveSurfer.js
- Backend: Python, Flask, Whisper, RoBERTa, PyTorch
- DevOps: Docker, Docker Compose
The following pull requests were created as part of the GSoC period:
The project successfully delivered a complete audio sentiment analysis workflow, allowing RUXAILAB to support deeper emotional insights during moderated usability tests.
- ✅ Audio region-based sentiment analysis.
- ✅ Seamless backend–frontend integration.
- ✅ Improved moderator experience with rich visual feedback.
- ✅ Scalable architecture for future sentiment-related features.
| Item | Description |
|---|---|
| Organization | RUXAILAB |
| Program | Google Summer of Code 2024 |
| Contributor | Basma Elhoseny |
| Project | Sentiment Analysis for Moderated Usability Tests (Audio) |
| Technologies | Python, Vue.js, Firebase, Whisper, RoBERTa, Docker |
| Topics | Sentiment Analysis, UX Research, Audio Processing, AI |
| Duration | May – August 2024 |
Special thanks to mentors Karine Pistili, Marc, and Vinícius Cavalcanti for their guidance, feedback, and continuous support throughout the GSoC journey.
Submitted as part of Google Summer of Code 2024 – Final Work Proof
© 2024 RUXAILAB • Developed by Basma Elhoseny