用于方面提取的多元关系协作学习模型

方面级情感分析广泛应用于商品评价、餐饮、电商决策等,该任务的一个核心点是方面词提取.目前常用方法是用观点词来辅助提取方面词对文本进行序列标注,或使用跨度标记法预测方面词开始与结束的位置.这些方法没有考虑到观点词提取、情感极性分类对方面词提取的影响.针对这个问题提出一种用于方面提取的多元关系协作学习模型,利用观点词提取、方面词提取、情感极性分类间的关系建模,在关系中实现多任务的协作学习与联合训练.在REST14、REST15和LAP14三个数据集上进行的实验结果表明,提出的方法优于目前的最新方法.

Multi-relationship collaborative learning model for aspect extraction

In recent years, aspect-level sentiment analysis has been widely used in product evaluation, catering, e-commerce decision-making, etc. A core point of this task is aspect word extraction. At present, the commonly used method is to use opinion words to assist in extracting aspect words to mark the text sequence, or to use span notation to predict the beginning and ending positions of aspect words. These methods do not consider the influence of opinion word extraction and emotion polarity classification on aspect word extraction. Aiming at this problem, this paper proposed a multi-relationship collaborative learning model for aspect extraction, which used opinion word extraction, aspect word extraction, and relationship modeling between emotion polarity classification to achieve multi-task collaborative learning and joint training in relationships. The experimental results on the three data sets of REST14, REST15 and LAP14 show that the proposed method is better than the current state-of-the-art method.