RestoPulse: AI-Powered Uzbek Restaurant Sentiment

• 

  Checking a single comment

• 

  aspects

• 

  checking a multiple comments

• 

  Comment review

🔴 Problem Statement Uzbekistan's hospitality and restaurant sector is growing rapidly. However, restaurant owners struggle to effectively monitor customer feedback. Manually analyzing hundreds of reviews across platforms like Google Maps, Telegram, and Yandex is time-consuming and subjective.

Existing global sentiment analysis APIs (like Google Cloud NLP or AWS Comprehend) lack robust support for the Uzbek language. They fail to understand local slang, spelling variations (e.g., mixing x and h), custom apostrophes (e.g., ’, ‘, ʻ, ʼ), and negation grammar (e.g., suffixing words with emas or yo'q). There is a clear need for a dedicated, localized business intelligence tool that can parse Uzbek customer sentiment and map it to specific operational areas.

🟢 Solution Overview RestoPulse is an AI-powered sentiment and aspect-based analysis system designed specifically for the Uzbek restaurant industry.

It processes customer feedback and provides two layers of intelligence:

Rating Classifier: Predicts a 1-to-5 star rating representing overall sentiment. Aspect Detector: Automatically maps comments to 6 key operational areas (Food Quality, Service, Pricing, Atmosphere, Cleanliness, and Location) and determines if the comment contains a complaint or praise for each area. RestoPulse solves the Uzbek language challenge using a hybrid engine: it combines a Multinomial Logistic Regression model trained on a custom-generated dataset with a rule-based linguistic correction layer that handles negation, typos, and apostrophe normalizations.

✨ Key Features Real-Time Sentiment Analysis: Type any Uzbek review and instantly get a 1-to-5 star classification, confidence percentage, and sentiment label. Aspect-Based Sentiment Analysis (ABSA): Detects specific feedback on Taom (Food), Xizmat (Service), Narx (Price), Muhit (Atmosphere), Tozalik (Cleanliness), and Joylashuv (Location). Interactive Data Visualizations: Custom Plotly charts showing star probability distributions and a diverging aspect bar chart representing positive vs. negative aspect sentiment. Batch CSV Analysis: Restaurant managers can upload a CSV export of customer reviews to instantly analyze trends, view overall metrics (Average Star Rating, Complaint/Praise distribution), and filter results. Integrated Data Scraper: A Selenium-based automation scraper to fetch real reviews directly from Google Maps without requiring expensive API keys. Robust Preprocessing: Built-in normalizer for Uzbek characters, common typos, and stop-word filtering. 🛠️ Technologies Used Programming Language: Python 3.x Machine Learning & NLP: Scikit-Learn (Logistic Regression, TF-IDF Vectorizer), NLTK/Regex, Langdetect User Interface & Viz: Streamlit, Plotly (Bar, Pie, and Diverging charts), Pandas, NumPy Web Scraping: Selenium WebDriver, Webdriver-Manager 👥 Target Users Restaurant Owners & Managers in Uzbekistan looking to improve food quality, service times, or hygiene based on data. Hospitality Business Analysts compiling market reports on dining trends. Food Bloggers & Aggregators looking to programmatically rate local dining spots.

Built With

• numpy
• pandas
• plotly
• python
• scikit-learn
• selenium
• streamlit