Multi-Task Learning for Sentiment Analysis with Hard-Sharing and Task Recognition Mechanisms

In the era of big data, multi-task learning has become one crucial technologies for sentiment analysis and classification. Most existing models are developed based on soft-sharing mechanism that less interference between different tasks than hard-sharing mechanism. However, there also fewer essential features model can extract with method, resulting in unsatisfactory classification performance. this paper, we propose a framework various fields. The is achieved by shared layer to build interrelationship among multiple tasks. Then, design task recognition reduce hard-shared feature space enhance correlation Experiments two real-world datasets show our approach achieves best results improves accuracy over methods significantly. training process enables unique representation space, providing new solution reducing analysis.