A Study On Topic-based Chinese Opinion Retrieval

The rapidly development and widely popularize of the Internet haschanged more and more people's lifestyle that people not only receiveknowledge from the Internet passively, but also communicate with the outerworld actively. The Internet is increasingly seen as an interactive media sothat lots of people likes to give their moods and comments on all kinds ofthings throw blog, BBS and other network media. It's important that how tofind and extract the opinioned information of a given topic from the vast dataof the Internet efficiently and rapidly. Opinion retrieval, which is a hotspot forthe research, is a document retrieval process, which requires documents to beretrieved and ranked according to their opinions about a query topic. Arelevant document must satisfy two criteria: relevant to the query topic, andcontains opinions about the query, no matter if they are positive or negative.The object of study of Chinese opinion retrieval are the opinions commentedby the Netizens, these opinions are valuable to the common Netizens, themanufacturers and the government. The opinions and reviews existed in theInternet is very important that Netizens can get references and suggestionsfrom these opinions, manufacturers can get the reviews for the product oftheir own and the competitors from the network opinion and the governmentalso needs to find out the popular feelings. So the research of Chinese opinionretrieval which aims to study on the network opinions is valuable for theresearch and many other applications.In this dissertation, the problems related to Chinese opinion retrieval areinvestigated, which includes query expansion, text retrieval and sentimentanalysis. we present an opinion retrieval algorithm that retrieval opinions for agiven topic according to the relevamce between the topic and its expansions, the topic and the sentiments and so on. It's based on the theory of textretrieval,sentiment analysis and the research for opinion retrieval of otherresearchers. The algorithm uses relevance to find the opinions and therelevance, which concerns the influence among all the elements of thisresearch theoretically, is a method to measure the affects between topics andother elements in the text such as its expansions, texts, the sentiments in thetexts and other elements.What's more, we study some algorithms which arepopular in English texts to find the Chinese opinions for ad hoc retrieval, weinvestigate the results of these methods with different arguments and differentresources.Then,we explorer the problems such as query expansion and theconstruction of sentiment dictionary. Besides, we use two different methods( sentiment classification and words-based fine-grained sentiment analysis ) toanalysis the sentiment orientation of the texts retrievaled by those algorithmswe discussed above. In the end , we experimentize all of the methods andalgorithms . The results show that the algorithm we presented is effective andgot a higher score than other methods .