Extracting Opinion Targets in a Single and Cross-Domain Setting with Conditional Random Fields

摘要: In this paper, we focus on the opinion target extraction as part of mining task. We model problem an information task, which address based Conditional Random Fields (CRF). As a baseline employ supervised algorithm by Zhuang et al. (2006), represents state-of-the-art employed data. evaluate algorithms comprehensively datasets from four different domains annotated with individual instances sentence level. Furthermore, investigate performance our CRF-based approach and in single- cross-domain setting. Our improves 0.077, 0.126, 0.071 0.178 regarding F-Measure single-domain domains. setting 0.409, 0.242, 0.294 0.343 over baseline.