Sentiment Analysis Using Word Sense Disambiguation

Sentiment analysis is crucial for many tasks and applications. For example, it is developed into a helpful tool for predicting market trends, monitoring social media, and being a reliable predictor of elections and products. For this task, numerous strategies were employed by researchers, ranging from rules-based strategies to machine learning and deep learning approaches to various data types. However, due to their failure to grasp the texts correctly, all of these techniques suffer from a lack of accuracy. In this work, an approach for extracting and using the (word sense disambiguation) of the word within the context is used as part of Initializing data and representing it. Two well-known algorithms were used (Bidirectional Long-Short Term Memory and Convolutional neural networks) for testing this approach in English. The system was evaluated using recall, precision, f-measure, and accuracy on the dataset (Amazon Kindle Store category). The results showed that the word sense disambiguation of the words is very useful, and the accuracy was increased by 4.21% for best cases and increased by 2% in worst cases. This work’s contribution can be summarized using word sense disambiguation (WSD), which aims to make the semantics of a word in context explicit by identifying the most appropriate meaning from a predefined sense, leading to many improvements in the performance of sentiment analysis.