Building A Multi-class XGBoost Model for Arabic Figurative Language

In the Natural Language Processing (NLP) field, the text classification becomes a task on which many scholars and researchers concentrate. Rhetorical methods in the Arabic language are among the means of linguistic expression that express opinions and feelings through written or spoken texts. It is essential to pay attention to this specialized research point in the Arabic language and in particular in the so-called Arabic rhetoric sciences that are concerned with figurative devices (i.e. simile, hyperbole and sarcasm). In this paper, we build the eXtreme Gradient Boosting (XGBoost) classifier to classify the multi-class Arabic figurative texts. The XGBoost is quite efficient for its speed and performance. The XGBoost classifier was developed, trained, and tested on this Arabic Figurative Corpus (AFC). The performance of the XGBoost classifier obtained as Fl-score is 88%.