Abdurahman-Developer/Local-Language-Complaint-Analyzer

An NLP system that automatically analyzes customer complaints for small businesses. Supports English, Swahili, and mixed-language input. # Local Language Complaint Analyzer **Author:** Abdirahman Mohamud Abdi | Reg No: BSCCS/2023/64091 ## Overview An NLP system that automatically analyzes customer complaints for small businesses. Supports English, Swahili, and mixed-language input. ## Features - Classifies complaints into: Product / Service / Delivery / Other - Detects sentiment: Positive / Negative / Neutral - Handles English + Swahili mixed text - Stores complaints in SQLite database - Live stats dashboard with category bar charts - Urgent (Negative) complaint flagging ## Tech Stack - Python 3 + Flask - scikit-learn (TF-IDF + Cosine Similarity) - SQLite (via built-in sqlite3) - Pure HTML/CSS/JS frontend (no build step) ## Setup ```bash pip install flask scikit-learn numpy python app.py # Open localhost ``` ## NLP Architecture 1. **Preprocessing** — lowercase, strip punctuation, remove stopwords (EN + Swahili) 2. **Classification** — TF-IDF vectorizer + cosine similarity against category keyword profiles 3. **Sentiment** — Lexicon-based scoring with negation detection ("not good" → Negative) 4. **Storage** — SQLite table: id, user_name, complaint_text, category, sentiment, timestamp ## Testing Examples | Complaint | Category | Sentiment | |-----------|----------|-----------| | "My order arrived very late" | Delivery | Negative | | "Your staff were very helpful" | Service | Positive | | "Product is broken and damaged" | Product | Negative | | "Bidhaa mbaya sana, imevunjika" | Product | Negative | | "Huduma mbaya, hawakusaidia" | Service | Negative |