All projects
AI & ML · Data ScienceJun 2023
Speech-based Movie Review Sentiment App
Speak a movie review and hear back how positive it is: speech-to-text, an LSTM sentiment model trained on IMDB reviews, and text-to-speech for the answer, served with Flask.
How it works
The app converts the user's spoken review to text, runs an LSTM model to score how positive it is, and reads the result back with text-to-speech. The backend is Flask.
The model
The notebook in the repo trains and evaluates the LSTM on IMDB movie reviews, with a confusion matrix, a classification report and a t-SNE view of how well the classes separate, and ends with the changes made for production.
Built with Ahmad Septian Brilliantino as a course team project.