Kyrgyz Headline Generator This project trains an mT5 model to generate headlines in Kyrgyz from article texts. It uses the Hugging Face transformers library.

Files: -KyrgyzHeadlineGeneration.ipynb - training and generation code -kg_dataset.csv - dataset with articles and headlines -Kyrgyz-Headline-Generation-using-Transformers.pptx — project presentation -“HeadlineGeneration”ProjectReport.pdf — project report

Setup Clone the repo and install dependencies: -pip install transformers datasets evaluate rouge_score nltk torch

Put your dataset path in the code (default is dataset.csv). Usage Run KyrgyzHeadlineGeneration.ipynb to train and evaluate the model. After training, use the saved model in headline_model/ to generate headlines without retraining. Example for inference: "from transformers import MT5ForConditionalGeneration, MT5Tokenizer tokenizer = MT5Tokenizer.from_pretrained("headline_model") model = MT5ForConditionalGeneration.from_pretrained("headline_model") input_text = "Your article text here" inputs = tokenizer(input_text, return_tensors="pt", truncation=True) outputs = model.generate(**inputs) print(tokenizer.decode(outputs[0], skip_special_tokens=True))"