Automatic text summarization has recently been an essential task in natural language processing (NLP). However, the development of summarizing systems needs datasets for proper evaluation. This requirement is necessary for less-resourced languages too. In this research, the first and free annotated corpus is produced and presented to evaluate abstract Kurdish text summarizing systems. News articles on this dataset have been utilized to collect the information. Also, an abstract Kurdish text summarization model based on the transformers has been developed for the first time to be evaluated by this dataset too. The current work can be a baseline for future research.
