Automatic Negation Detection in Arabic Reviews Using Supervised Classification Approach

Abstract

Sentiment Analysis (SA) is the process of determining the opinion expressed in a text written in a natural language, categorizing it as positive, negative, or neutral towards specific targets such as individuals, events, topics, products, organizations, or services. SA encounters various obstacles, one of which is the presence of negation. Negation, a linguistic phenomenon inherent in natural language, functions by reversing the meanings of sentences. It transforms positive statements into negative ones, impacting the polarity and sentiment conveyed in text. Detecting negation presents a notable challenge, especially within the context of the Arabic language, which is characterized by its intricate and multifaceted morphology. The ability to identify negation holds significant importance in the field of SA as it enhances the performance of various Natural Language Processing (NLP) applications. In this paper, we introduce an approach for the automated detection of negation in user-generated Arabic hotel reviews. We employ multiple supervised classification techniques, including Naïve Bayes (NB), Random Forest (RF), Logistic Regression (LogR), Support Vector Machines (SVM), and Deep Learning (DL), to analyse lexical and structural features extracted from the corpus. The results of our experiments have yielded promising outcomes, demonstrating the feasibility of our approach for practical applications. The classifiers exhibited highly comparable performance in identifying negation, with only marginal deviations in their performance metrics. Particularly noteworthy, the DL classifier consistently emerged as the top performer, achieving an exceptionally high overall accuracy rate of 99.24%surpassing established benchmarks in Arabic text processing and underscoring its potential for practical applications. These findings hold significant implications within the realm of Arabic text processing.