Case Study

Sentiment Analysis
with Audio Input

An interactive interface that detects sentiment in comments about pop artists, based on audio recordings.

Real-time
Audio Analysis
100%
Immediate Feedback
Python| Openai-whisper, TF-IDF & Streamlit
Technologies and Libraries

Project Journey

From concept to implemented solution

Início

The Challenge

Create an interactive interface that detects sentiment in comments about pop artists, based on audio recordings.

  • Detect positive or negative sentiment
  • Transform audio into text for analysis
  • Provide quick feedback to the user
  • Maintain a simple and intuitive interface
Planning

Solution Architecture

We used Streamlit for the interactive interface, Whisper for audio transcription, and Machine Learning models for sentiment analysis.

  • Streamlit for front-end
  • OpenAI Whisper for audio transcription
  • Scikit-learn for sentiment classification
  • NLP for text preprocessing
  • Real-time feedback to the user
Development

Audio Processing

We implemented audio recording in the browser and direct conversion to numpy array, avoiding temporary files.

  • Audio normalization and resampling
  • Transcription via Whisper in memory
  • Preprocessing of the transcribed text
  • Sentiment classification
Testing

Testing and Feedback

We conducted tests to ensure model accuracy and a smooth user experience in the app.

  • Compare prediction with real sentiment
  • Interface testing with users
  • Transcription time optimization
  • Bug fixes and UX improvements
Result

Project Success

The final app allows any user to record comments and receive instant sentiment analysis, with an intuitive interface.

  • Real-time feedback
  • Accurate audio transcription
  • Efficient sentiment classification
  • Scalable project for more artists or languages

See the repository

Interactive Application

Try the sentiment analysis app

Open Pop Artist Sentiment Analyzer