pdf:https://www.aclweb.org/anthology/2020.acl-main.290.pdf
特点:
- 简单的无监督的方面提取方法
- 只需要利用POS标注和领域词向量
- 在小规模数据级上训练
- 模型自动进行分配方面标签。
模型示意图:
方法:
step 1:aspect term extraction. use spaCy for tokenization and POS tagging
step 2: aspect selection using Contrastive Attention.
(rbf, Radial Basis Function)
S: a sequence of word embeddings
A: aspect representation vectors
Step 3: assigning aspect labels.
C:the set of aspect labels. 计算时采用方面标签的词向量。
分析:
- The in-domain word embeddings are required to achieve good performance.
- Several of the errors relate to Out-of-Vocabulary (OOV) or low-frequency items,
- Because restricting to nouns, the model also misses aspects expressed in verbs