Text Aspect-level Sentiment Analysis based on Multi- task Joint Learning

Text sentiment analysis is an important research topic in the field of natural language processing. Compared with coarse-grained text sentiment analysis with a single sentiment tendency based on sentences or documents, fine-grained aspect- level sentiment analysis is more suitable for practical application scenarios and becomes more difficult at the same time. In this paper, we propose a multi-task-based joint learning model for aspect-level sentiment analysis of text, using the BERT_CBiGRU composite network with stronger text semantic representation as the main learning task for aspect-level sentiment analysis, and designing an auxiliary learning task for aspect target identification. For the sample imbalance problem, we introduce the focal loss function Focal Loss. Final experiments show that our model has a certain improvement compared with previous models.