ArEmotive Bridging the Gap: Automatic Ontology Augmentation Using Zero-Shot Classification for Fine-Grained Sentiment Analysis of Arabic Text

Abstract

Human-computer interaction remains one of the final frontiers to conquer while held in perspective with the rapid developments and technology growth over recent years. It is an arduous task to convey the true human intent to the machine in order to generate a computerized relevant decision in a certain field. In recent years, focus has shifted to cover fields of study that relate to Sentiment Analysis (SA) to improve and ease the tasks of our daily lives. We Propose ArEmotive (Arabic Emotive), a fine-grained sentiment analysis system that is human-independent which can automatically grow its source of information allowing for more precision and a greater dataset each time it is used through ontology augmentation and classification. Our proposed architecture relies on multiple data sources running through certain pipelines to generate a central online repository utilized by any mobile system to access this info-base. This system is important because many researchers in the field of automated ontology alignment and ontology mapping achieved a semi-automated approach to map new ontologies out of old ones or to extend already existing ontologies with data from new ones. ArEmotive identifies fine-grained emotions in text based on a dynamic ontology enriched through ontology alignment, mapping and machine learning assisted classification, resulting in a structure that contributes in: a centralized dataset ever growing to fit the need of the users, a sustainable structure able to allocate new data sources without the need to modify the system, ability to generate appropriate information even with the absence of 'parent' sources.