A Neural Group-wise Sentiment Analysis Model with Data Sparsity Awareness

Sentiment analysis on user-generated content has achieved notable progress by introducing user information to consider each individual’s preference and language usage. However, most existing approaches ignore the data sparsity problem, where of some users is limited model fails capture discriminative features users. To address this issue, we hypothesize that could be grouped together based their rating biases as well degree consistency knowledge learned from groups employed analyze with data. Therefore, in paper, a neural group-wise sentiment awareness proposed. The user-centred document representations are generated incorporating group-based encoder. Furthermore, multi-task learning framework jointly modelusers’ consistency. One task vanilla populationlevel other groupwise analysis. Experimental results three real-world datasets show proposed approach outperforms state-of the-art methods. Moreover, case study demonstrate its effectiveness modeling variances.