摘要：We provide a contribution that points up the potential use of frame semantics and FrameNet inparticular for sentiment analysis. We address several key problems in current sentiment analysis, which ischaracterized by shallow approaches, pragmatic focus, and ad-hoc creation of data sets and methods. We arguethat progress towards deep analysis depends on a) enriching shallow representations with linguisticallymotivated, rich information, and b) focusing different branches of research and combining resources and workforces to join hands with related work in NLP. We propose SentiFrameNet, an extension to FrameNet, as a novelrepresentation for sentiment analysis that is tailored to these aims
