Subjective text recognition is the premise of emotion computation. The current method is the dictionary-based method and statistical-based method, while ignoring the subjective clues which contain rich emotional information and the accuracy is not high. To solve this problem, this paper selects the associated word, the emotional word as well as the indicative verb, the interjection, the degree adverb, with sentiment of punctuation six kinds of subjective clues ingredients as the subjective text recognition basis, establishes subjective clues character dictionary, then identifies the subjective text with the Naïve Bayes classifier. The experimental results show that this method proposed this paper has higher accuracy.