The Constitution of a Fine-Grained Opinion Annotated Corpus on Weibo

Sentiment analysis on social media represented by Weibo is one of the hotspot research problems in NLP. A comprehensive and systematic fine-grained annotated corpus plays a significance role. In this paper, considering the characteristics of Weibo, we focus on the constitution of a fine-grained, hierarchical opinion annotated corpus and design a set of labelling specification. We manually annotate the opinion sentences with a part of ones containing hidden opinion which can be useful for implicit sentiment analysis. Then a fine-grained aspect extraction, namely opinion triples like is finished for aspect-level sentiment research. Moreover, we establish an evaluation method for the task of fine-grained aspect extraction which has been applied in evaluation for years. The corpus was used in the task of COAE2015, and it will be a useful resource for the related research on social media sentiment analysis.