How aakinolaj/reviewers-dashboard works
Project type and purpose
• This is a single-page NLP dashboard built on Streamlit (a Python web-app framework).
File structure
• Root:
– luxor_dashboard_app.py → the main Streamlit application
– luxor_reviews_only.csv → sample data
– requirements.txt → dependency list
• No separate frontend/backend directories because Streamlit bundles UI and backend logic in one process.
Key insights from README
• Describes core components: Streamlit UI, topic modeling, sentiment analysis, interactive charts.
• Mentions deployment on Streamlit Cloud.
Main components
a. User Interface Layer
• Browser clients (users interact via web browser)
• Streamlit’s frontend (widgets, upload button, chart panels)
b. Application Logic Layer
• Streamlit server (luxor_dashboard_app.py)
• Request handler: receives uploaded CSV
• Data pre-processing module (e.g., tokenization, cleaning)
c. NLP Processing Modules
• Gensim LDA engine for topic modeling
• TextBlob for sentiment analysis (requires corpora)
d. Visualization Layer
• Plotly chart generator (bar charts, scatter plots)
• Download module to export annotated CSV
e. External/Infrastructure Services
• Streamlit Cloud (CI/CD: GitHub → build container → deploy)
• GitHub repository (source code + docs)
• (Optionally) TextBlob’s online corpus downloader
Relationships & data flow
1. User → “Upload CSV” widget
2. Uploaded file → Streamlit server
3. Streamlit server → Pre-processing → NLP modules (LDA, TextBlob)
4. NLP results → Plotly generator → back to browser for rendering
5. Export button → Streamlit server → CSV download to user
6. Deployment pipeline: GitHub → Streamlit Cloud → Streamlit server
Architectural patterns & design principles
• Single-process, event-driven UI (Streamlit)
• Modular NLP pipeline: separate concern for data ingestion, modeling, and visualization
• Stateless per user session (each upload runs fresh analysis)
• Use of third-party libraries rather than custom implementations