Research On Key Issues In Text Orientation Analysis

Today, China's e-commerce has been extremely popular, Taobao, Jingdong and other large shopping sites have accounted for most of the market. Companies need to understand the user's attitude towards the product in order to gain business benefits and consumers to decide. The use of people to mark text emotion, laborious and time-consuming, it is necessary to use the computer to achieve automatic the emotional tendency of text. This kind of technology is called text orientation analysis. Now, the technology has achieved many research results. This paper mainly studies the problems in the methods of text orientation analysis.In the text orientation analysis based on machine learning, the key research is the problem of low accuracy caused by the test and training text not in the same field. Aiming at the feature dimension reduction, an feature clustering model is proposed, which is based on generic domain framework. In calculation method of cross domain sentiment analysis based on weighted simrank, the total weighted failed to consider the synonym when feature alignment, and feature clustering is used. Related experiments show that, ensuring the accuracy rate, meanwhile saving the memory space, aleviating the data sparse problem.In the text orientation analysis based semantic, it focuses on word orientation. In view of the commonly used method of word tendency calculation, over reliance on the repository and the semantic nexus can not be excavated. This paper proposes a method to calculate the orientation of emotional words based on word vector. This method is based on the object oriented domain emotional words, which is exited in the specific field and has obvious emotion. Based on Google's Word2vec, word vector is obtained through neural network. The cosine distance between vectors is used as the words' similarity, which judges the word and the reference words' similarity degree, and then judges the emotional tendency of the word. Experiments show that the method has the field flexibility and high accuracy.