Lingerew-10/Amharic_Intention_Detection: Pragmatic Intention Detection in Amharic using transformer-based model

This release provides the implementation and resources accompanying the research on Pragmatic Intention Detection in Amharic using a Transformer-Based Model. Contents Amharic intention detection dataset Dataset files and supporting resources Transformer-based model implementation Training and evaluation code Experimental results Documentation and usage instructions Model The project uses AfroXLM-R (Davlan/afro-xlmr-base), a multilingual transformer model with support for African languages, for Amharic pragmatic intention classification. Dataset The dataset contains 11,475 Amharic text samples organized into 9 pragmatic intention classes, with separate training, validation, and test splits. Evaluation The trained model achieves 83.33% test accuracy and a 0.8222 Macro F1-score on the held-out test set. Reproducibility The repository provides the necessary datasets, code, and documentation to facilitate reproduction of the reported experiments. Citation If you use this dataset, code, or research in your work, please cite the associated research publication. Version: v1.0.0 Status: Stable research release