Arabic aspect based sentiment analysis using bidirectional GRU based models

Aspect-based Sentiment analysis (ABSA) accomplishes a fine-grained that defines the aspects of given document or sentence and sentiments conveyed regarding each aspect. This level is most detailed version capable exploring nuanced viewpoints reviews. The bulk study in ABSA focuses on English with very little work available Arabic. Most previous Arabic has been based regular methods machine learning mainly depends group rare resources tools for analyzing processing content such as lexicons, but lack those presents another challenge. In order to address these challenges, Deep Learning (DL)-based are proposed using two models Gated Recurrent Units (GRU) neural networks ABSA. first DL model takes advantage word character representations by combining bidirectional GRU, Convolutional Neural Network (CNN), Conditional Random Field (CRF) making up (BGRU-CNN-CRF) extract main opinionated (OTE). second an interactive attention network GRU (IAN-BGRU) identify sentiment polarity toward extracted aspects. We evaluated our benchmarked hotel reviews dataset. results indicate better than baseline research both tasks having 39.7% enhancement F1-score opinion target extraction (T2) 7.58% accuracy aspect-based classification (T3). Achieving F1 score 70.67% T2, 83.98% T3.