The Research And Implementation On The Key Technology Of Internet Public Opinion Analysis

More and more people are influenced by internet as the amount of internet users increases explosively. However, if there is inefficient management of internet, unreal talking and malevolent exaggeration on some sensitive cases and emergencies could mislead and cheat people, then enlarge people's unsatisfactory, at last impact and break the stability and harmony of society. The fitness of internet information has drawn great attention from all levels of governments and it is necessary to effectively supervise the topics and expressions on ideology safety. The webpage on internet are increasing exponentially every day and it is impossible to screen and analyze all the information on each webpage by manpower. The only option to establish overall, effective and fast monitoring and early warning mechanism of public opinion is that to adopt automatic computer technology so that the internet can develop in a fast and healthy way. Therefore, the study of gaining and analyzing technology of internet public opinion has been an urgent and import issue.This paper deeply studied the key technology of text semantic orientation. It analyzed the advantage and disadvantage of present semantic orientation identification technology and then integrated the good performance of Hidden Markov Model on text processing, applied text semantic orientation analyzing method studied in this paper into internet public opinion analyzing system to attain the analyzing and early warning of public opinion.The purpose of text semantic orientation analyzing is to judge the sentiment tendency of text towards the evaluated objectives. The tendency is supportive, opposed or neutralized? Similar comments must have some text with inherent relationship. As a variety of ways the performance of internet public opinion, this paper put internet comments as its study objective and tried to spread Hidden Markov Model from the field of model identification in which it was already successfully applied to the system of semantic orientation analyzing. The difference from traditional orientation identification system is that this theory put unknown text in sequent state through building Hidden Markov classified Model and then got the tendency from all the tendentious words and eventually chose the tendency of most words as the overall semantic orientation of the text.In the experimental system of this paper, we use the integrated development environment of Myeclipse7.0 platform, consist of three modules:the corpus collection, building models and semantic recognition. The corpus collection module provides data support for the other two modules. The building models module trains the corpus that collected by the corpus collection module, and then obtains the semantic recognition model. The semantic recognition module completes the specified text semantics orientation recognition. In this paper, we conduct closed test and open test respectively on the data from Tencent News Forum, and the results show that this analyzing model can recognize the semantic orientation of various kinds of unknown text every well both in open and close tests. What's more, the chance of recognizing will be much more stable and higher when the data of practicing is larger and more overall.