The Sentimental Orientation Analysis Of Sentence Based On Sentiment Dictionary

The sentimental orientation analysis of Chinese texts refers to extract and analyze subjective information from texts, including judging the sentiment polarity and mining each relevant element of sentimental orientation discussion which contains opinion object, the sentimental orientation of texts (including positivity, negativity, neutrality), and knowledge of how the texts related to the orientation intensity. With the more popularity of the Internet, come out the more commentary texts which are needed to be analyzed automatically, which makes a hot research area the sentimental orientation analysis of Chinese texts. Therefore, to determine the sentence-level sentimental orientation of Chinese texts is of a very basic and important research topic.In this paper, some relevant information and typical sentimental orientation analysis algorithms are firstly and carefully studied, the principle and shortage of the common methods are then discussed, including some problems in the process of sentiment analysis. Finally, aim to solve these problems the corresponding algorithms are proposed and the relevant experiments are made to compare with previous methods.In this paper the following four aspects are concerned as below:1. The processing algorithm is proposed to how to ensure the balance of training data when the size of training corpus is serious imbalance; by this way, the negative influence can be reduced resulted from the imbalance of training data size.2. The algorithm of splitting the large-scale sentimental words dictionary into subs is proposed through studying the confidence evaluation methods of sentiment words to lower the negativity that is brought by classifying the sentiment polarity with low confidence sentiment words, and the related experiments are done to verify the effectiveness of the proposed method.3. The algorithm of establishing the rule set is proposed by deeply a nalyzing the training corpus to solve how to correctly analyze the sentime nt orientation problems resulted from the shortage of parts of sentimental words or from the weak ineffective emotional words contained in the testi ng corpus.4. A multi-level classification algorithm is proposed to work out the problem that by using only one method cannot give consideration to the precision and the recall. Of multi-level classification algorithm, the neutrality and the polarity of the subjective sentences are determined firstly, the polar sentences are then divided into positivity and negativity. In the process of binary polarity classification, a different strategy is adopted that the result can only be obtained from one layer of multi-level classification model. The experiments indicate that both the precision and the recall are improved.