Medical Sentiment Analysis Dataset for Kurdish Short Text over Social Media

Data Description

The Kurdish language is one of the languages of the Middle East that is used for speaking by Kurdish people. Central Kurdish (Sorani) and Kurmanji are two popular dialects of the Kurdish Language [1, 2]. In this project, the Sorani dialect is used for collecting the database. The texts of Sorani Kurdish are like Persian, Arabic, and Urdu that are written from right to left. The characters among those languages are almost like each other but sometimes have different Unicode as shown in Table 1. The number of Sorani Kurdish alphabets is 36 that is divided into vowels (ا, ه‌, و, ۆ, وو, ى, ێ) and consonants (ئ, ب, پ, ت, ج, چ, ح, خ, د, ر, ڕ, ز, ژ, س, ش, ع, غ, ف, ڤ, ق, ک گ, ل, ڵ, م, ن, ه, (و, ى)) [3, 4]. The character (و, ى) used as vowels and constants based on the positions of the word for example the word (گوڵ) (gull) means (Flower) the (و) is vowel while the word (وازى) (wazi) means (game) the (و) is constant and the word (يارى) (yari) means (play) the first (ى) is constant by contrast the second one is a vowel. The Kurdish language is complex and has different scripts (no standard) for Sorani dialect for example in some sources (ك) is used instead of (ک) [3, 4].

Dataset Labeling:

After collecting the dataset, another important step is labeling the samples. For this purpose, three annotators read the samples accurately and manually labeled the unlabeled samples for two classes (medical and non-medical). This process needs a huge effort and consumes time. For labeling each sample, the annotator annotates the sample based on some special words in the medical domain and the meaning of each sentence as shown in Table 4: