Sentiment Analysis in Saudi Arabic Dialect for HAJJ Season Services Using Twitter Data

Recent years have seen a lot of interest in sentiment analysis. However, many failed attempts have been made to develop a Sentiment Analysis model for morphologically rich languages such as the Arabic dialect. We aimed to fill the gap in this study by creating a new manual corpus of the Saudi Arabic dialect and a novel polarity lexicon. The corpus collects tweets from various services (Transportation, Hotel, Restaurant) annotated on a sentence level by a developed online platform. We also discussed some automated, online Sentiment Analysis tagging tools that could be used to explore all annotated data. Furthermore, we investigated the impact of different pre-processing techniques on the Saudi Arabic dialect process. Finally, we presented a new method for the Saudi dialect that produced an excellent classification result, with an 85% accuracy for the Support Vector Machine and an 88% precision for Nave Bayes.