Tourism-Related Opinion Mining

Tourism-Related Opinion Mining Chuan-Jie Lin Pin-Hsien Chao Department of Computer Science and Engineering National Taiwan Ocean University {cjlin, m96570038}@ntou.edu.tw Abstract This paper focuses on the tourism-related opinion mining, including tourism-related opinion detection and tourist attraction target identification. The experimental data are blog articles labeled as in the domestic tourism category in a blogspace. Annotators were asked to annotate the opinion polarity and the opinion target for every sentence. Different strategies and features have been proposed to identify opinion targets, including tourist attraction keywords, coreferential expressions, tourism-related opinion words, a 2-level classifier, and so on. We used machine learning methods to train classifiers for tourism-related opinion mining. A retraining mechanism was proposed to obtain the system decisions of preceding sentences as a new feature. The precision and recall scores of tourism-related opinion detection were 55.98% and 59.30%, respectively, and the scores of tourist attraction target identification among known tourism-related opinionated sentences were 90.06% and 89.91%, respectively. The overall precision and recall scores were 51.30% and 54.21%, respectively. Keywords: tourism-related opinion mining, tourist attraction target identification, opinion analysis 1 Introduction The blogspace is a large resource for opinion mining. Opinion extraction methods are valuable for a wide range of applications. Our initial interest is to extract opinions related to tourist attractions from blog articles. When planning a tour, it is helpful to see other people’s opinions about the tourist attractions. However, two issues arise when trying to apply published methods to retrieve opinions of tourist attractions: (1) Sentence-level or document-level: A travel article is often multi-topic because a travel route often includes several tourist attractions. Therefore, the opinion analysis for a specific tourist attraction should be carried out in sentence level, not in document level. (2) Opinion topic or opinion target: Tourist attractions may be treated as topics (queries in IR) or targets of opinions. Consider the following two sentences selected and adapted from our dataset: The Dream Lake is a beautiful place. The water is green and clear. Both kinds of sentences are considered as tourism-related opinions by us. However, 3 Proceedings of the 22nd Conference on Computational Linguistics and Speech Processing (ROCLING 2010), Pages 3-16, Puli, Nantou,Taiwan, September 2010. their opinion targets are not the same. The opinion target of the first sentence is “the Dream Lake” itself, while the target of the second sentence is “the water (in the Dream Lake)”. Both sentences are related to the same topic “the Dream Lake”, but the second sentence does not contain its topic. We find difficulties to apply known-developed methods due to these reasons discussed here. Opinion mining and analysis have been widely studied in several topics, including opinion detection and polarity classification [1] [2] [3] [4], opinion holder finding [5] [6] [7], and opinion summarization [8]. Opinion retrieval is one of the research topics relevant to our work. Godbole et al. [9] estimated the polarity scores for a large set of named entities. But they measured the scores by the co-occurrences of named entities and opinion words, so the opinionated …

Content truncated to 500 words