Transformer Based Amharic Headline Generation using Sub-word2Vec Representation

The major component of a news item that aids in giving a quick peek at the article's theme is the headline. The process of creating a headline involves summarizing an article to produce a summary that is sufficient in terms of clarity and consistency. Although many different types of research have been done in the domain of headline generation, it is still difficult to create an effective machine-generated title. Additionally, those methods do not work well for morphologically rich languages like Amharic. We proposed a neural headline generation model using transformers and different types of embedding techniques.