Three-level binary tree structure for sentiment classification in Arabic text
Hajar Ait Addi, Redouane Ezzahir, Abdelhak Mahmoudi · 2020
The advent of web 2.0 platforms allowed users to generate and share textual content. This results in an explosive increase of online personal opinion. Sentiment Analysis, which is a recent field of Natural Language Processing, aims to predict the orientation of sentiment present on this massive textual data. This plays a vital role in many applications, such as recommender systems, customer intelligence, information retrieval and psychological study of crowd. Most existing approaches in sentiment analysis trait only positive, negative and neutral classes, ignoring the class strength (weak or strong positive/negative). In this paper, we propose an innovative approach for multi-class hierarchical sentiment classification in Arabic text based on a three-level binary tree structure. Experimental results show that our approach gives significant improvements over other classification methods.