Multi-document Sentiment Abstraction Study For Comment Texts

At present, people can make comments on the current news events or shopping goods anytime and in any place as they like. The comment text usually contains the reviewers' sentiment information, which can reflect that the event or goods is good or bad. But there are such a large number of redundant comment texts in the network that sentiment comment texts can not be extracted or found easily. In order to keep users from useless information, and solve the problem of interdisciplinary sentiment analysis and topic content extraction, this paper utilizes sentiment summarization technology to analyze, process and integrate the comment text, finally, get the sentiment summarization the user wanted.At first, this paper studies and discusses the key technology of processing the comment text. Then, by use of sentiment dictionary construction method, the domain sentiment dictionary is constructed for text sentiment analysis. At last, the improved sentiment key sentence extraction algorithm is applied to the multi-document summarization extraction. In this paper, the work mainly includes the following three aspects:(1) Put forward a kind of domain sentiment lexicon construction based on corpus, then research on the sentiment analysisComment text usually contains subjective emotion and they are distributed in many domain. Therefore, we have to correctly analyze the text sentiment information before conducting sentiment summarization. The integrity of current sentiment dictionary is the foundation of sentiment analysis. By fully combined with other emotional dictionary and domain corpus, this paper proposes a method of building domain emotional dictionary based on corpus.This method firstly uses the domain corpus whose emotional tendency has been known, extracts the evaluation objects of the domain corpus, calculates the correlation between the evaluation objects and emotional words in other basic dictionary, extracts words that has stronger correlation, calculate their orientation value according to their importance degree in corpus, finally forms a new sentiment dictionary.(2)Do research on multi-document sentiment summarization methods, put forward a kind of multi-document sentiment summarization method based on key sentiment sentence extractionBefore sentiment summarization, firstly we judge and analyze key sentiment sentence in the text. Because the key sentiment sentence can not only express the theme of the comment text, but also can express people's subjective views, we apply the improved key sentiment sentence extraction method to multi-document summarization, propose a kind of multi-document summarization method based on key sentiment sentence extractionThis method firstly calculate the topic relevance, keyword attributes of the sentences in the text, uses domain emotional dictionary to analyze the emotional attributes of the sentence, and gives different weights for each attribute. The emotional attribute's weight is bigger. Finally, summarize them to calculate the score of the sentence, extract the sentences that are with high score according to the maximum edge related redundant emanation algorithm to form the final summarization. Experimental results show that the extracted sentiment summarization by our method can match well with that extracted by experts.(3)Design and implement the prototype system of multi-document sentiment summarization based on key sentiment sentence extractionIn the process of multi-document sentiment summarization extraction, we design the corresponding function module for each aspect, and implement the system of multi-document sentiment summarization based on key sentiment sentence extraction.The system can mine and extract, analyze and process the comment texts in the network, provides users with intuitive sentiment summarization. The shown summarization can not only express the theme of comment, also grasp the corresponding emotional information.