Medical Dataset Classification for Kurdish Short Text over Social Media
This work addresses text classification for Kurdish language in medical contexts, but it is incremental as it focuses on dataset creation and basic preprocessing without novel methodological advancements.
The authors tackled the problem of classifying short Kurdish text from social media as medical or non-medical by creating a dataset of 6,756 comments and applying preprocessing techniques, achieving a dataset with 45% positive (medical) and 55% negative (non-medical) classes.
The Facebook application is used as a resource for collecting the comments of this dataset, The dataset consists of 6756 comments to create a Medical Kurdish Dataset (MKD). The samples are comments of users, which are gathered from different posts of pages (Medical, News, Economy, Education, and Sport). Six steps as a preprocessing technique are performed on the raw dataset to clean and remove noise in the comments by replacing characters. The comments (short text) are labeled for positive class (medical comment) and negative class (non-medical comment) as text classification. The percentage ratio of the negative class is 55% while the positive class is 45%.