Supervised Domain Adaptation via Label Alignment for Opinion Expression Extraction.
CICLing(2017)
摘要
In this paper, we propose a supervised domain adaptation technique for opinion expression extraction task. The technique generates low dimensional projections that can improve the performance of a sequence model (e.g. CRF) in the target domain by align features with the true label sequence. We test our methods on product reviews and observe significant improvement in performance in comparison to baseline methods.
更多查看译文
关键词
Domain adaptation, Sequence labeling Sequence extraction
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络