Modeling and extracting evaluation objects on social media short content

With the development on newly social media such as micro-blog, people usually spread sentiments on various kinds of topics, and the social media has become an effective platform to mine the public opinions. As this kind of big data has the characteristic of comments diversity, it is necessary to extract useful information such as sentiment or emotion from the big data. But the traditional text mining algorithm does not fit the need with this kind of short and fractional content. This paper proposes a sentiment evaluation object extraction approach based on lexical & syntactic rules. We combine the lexical analysis and dependency parsing together to extract the corresponding evaluation objects and find that the lexical rules have better performance on this kind of short content corpus. Experiment results show the feasible of the approach, and the existing problems and further works are also presented in the end.