How mohtadaayesha/critix works
Critix is a Gradio movie-review analyzer. A user submits a review; the app predicts sentiment, compares the review with sentiment-matched community reviews to identify recurring aspects, and independently retrieves movie recommendations. It renders sentiment, community verdict, and recommendation cards. The graph separates the UI, analysis, and retrieval capabilities. The sole source file was sampled; asset-loading relationships shown below are explicit in that source, while no additional implementation modules are present in the tree.
Main entry points: app.py Review pipeline · sentiment_lstm_model.keras Sentiment model · tokenizer.pkl Text tokenizer · index_positive.faiss Positive review index
Components 18
User Interface
- Gradio UI
app.py - Review pipeline
app.py - Result cards
app.py
Review Analysis
- Sentiment prediction
app.py - Sentiment model
sentiment_lstm_model.keras - Text tokenizer
tokenizer.pkl - Community retrieval
app.py - Aspect extraction
app.py - Critic verdict
app.py
Similarity & Movies
- Text embedder
app.py - Positive review index
index_positive.faiss - Negative review index
index_negative.faiss - Positive review texts
texts_positive.csv - Negative review texts
texts_negative.csv - Movie retrieval
app.py - Movie index
movie_index.faiss - Movie metadata
movie_metadata.csv
Other components
- Reviewer
Connections 19
- Reviewer → Gradio UI: submits review
- Gradio UI → Review pipeline: dispatches
- Review pipeline → Sentiment prediction: predicts sentiment
- Sentiment prediction → Text tokenizer: tokenizes with
- Sentiment prediction → Sentiment model: predicts with
- Review pipeline → Community retrieval: retrieves examples
- Community retrieval → Text embedder: encodes review
- Community retrieval → Positive review index: searches by sentiment
- Community retrieval → Negative review index: searches by sentiment
- Community retrieval → Positive review texts: selects matching texts
- Community retrieval → Negative review texts: selects matching texts
- Review pipeline → Aspect extraction: extracts aspects
- Review pipeline → Critic verdict: builds verdict
- Review pipeline → Movie retrieval: retrieves movies
- Movie retrieval → Text embedder: encodes review
- Movie retrieval → Movie index: searches
- Movie retrieval → Movie metadata: selects movie details
- Review pipeline → Result cards: renders results
- Result cards → Gradio UI: returns cards