Academic Editor: Youssef EL FOUTAYENI
Received |
Accepted |
Published |
31 January 2020 |
15 February 2020 |
10 March 2020 |
Abstract: The development of social media allows to increase the textual information which provides different opinions. So, Sentiment Analysis (SA) of Arabic text documents have become a real challenge in the field of Natural language processing (NLP) [1]. This paper consists to identify automatically the opinion into positive/negative. This task required two phases: the first is preprocessing phase which applied several steps including normalization, tokenization, removing the stop words and stemming [2], while the second is Feature selection (FS) using grey wolf optimizer (GWO) [3]. The transformation of textual data to numerical data is realized by computing the TF-IDF matrix which increase the dimensionality of the space research. Thus, the process of FS is obligatory in the field of NLP in order to select the relevant words using a wrapper GWO based FS. The experimental results are assessed on two public datasets: AJGT and OSAC. The wrapper GWO employed K-nearest neighbors as classifier shown a good performance for opinions classification with 80%.
