birhe23/Amharic_idiom_classification

# Amharic Idiom Classification This repository contains an Amharic idiom classification project that combines a Flask-based web dashboard with text preprocessing, model training, and model evaluation. ## Project Overview - A Flask application for browsing an Amharic idiom dataset, filtering idioms, and predicting idiom sentiment labels. - A `sklearn` pipeline model using `TfidfVectorizer` and `LogisticRegression` saved as `classifier_model.joblib`. - A labeled Amharic idiom dataset in `labeled_idiom.csv` and a results summary in `model_results.json`. - Experimental Jupyter notebooks for idiom classification without and with word embedding approaches. ## Current Status - `app.py` is a working Flask web app with: - `/login` authentication page (dummy credentials: `admin` / `admin123`) - `/dashboard` dataset summary and idiom browsing UI - `/predict` prediction endpoint for new idiom text - `/download` CSV export of the current filtered idiom view - `classifier_model.joblib` is built from `labeled_idiom.csv`. - `model_results.json` contains the current evaluation metrics: - Accuracy: `0.92` - Precision: `0.90` - Recall: `0.88` - F1 score: `0.89` ## Repository Contents - `app.py` - Flask application implementing login, dashboard, prediction, filtering, and download. - `classifier_model.joblib` - saved trained model used by the app. - `labeled_idiom.csv` - labeled Amharic idiom dataset used for training and dashboard display. - `model_results.json` - JSON file with evaluation metrics. - `README.md` - project documentation. - `tests/test_app.py` - unit tests for the Flask app routes and basic functionality. - `Amharic_idiom_dictionary.txt` - dictionary resource file for Amharic idiom preprocessing. - `idiom_and_meaning.csv` and `Amharic_idiom_preprocessing.ipynb` - dataset and preprocessing notebook resources. - `idiom_classification_without word embedding.ipynb` - notebook exploring classification without pretrained embeddings. - `idiom_classification_using word e …