ABSTRACT
The above methods reveal the application rules of languages from different perspectives, i.e. semantic relationship reveals the universal rules of human languages, representing priori knowledge; statistical relationship reveals the application habits of languages in specific fields, representing evidences. However, the analysis results might contradict with the actual situations because some words will have different meanings in specific fields or the context due to the ambiguity of words and non-standard Internet language. A lexicons that is constructed based on massive corpus statistics will reflect customs of language expressions, but there are many undesired results within it and it cannot cover all related emotion words.
