Abstract: The Internet has driven the development of tourismrelated industries， and homestays are also favored by more and more tourists because of their characteristics. How to fully and effectively analyze the emotional polarity in online reviews， explore its potential value， and help managers better improve the homestay experience has become a research hotspot. Therefore， this paper takes Guilin Yangshuo homestay reviews， uses octopus crawler software to crawl data from Ctrip and Qunar travel websites， and uses Dalian polytechnic emotional vocabulary ontology database as the basic emotional dictionary， adds screened emotional words in the homestay field， and constructs an emotional dictionary in the homestay field. Compared with the basic dictionary， the accuracy rate of the domain dictionary constructed in this paper is improved by 2.7% in positive commentary， and the F value is increased by 7.4%. The accuracy rate of negative reviews was improved by 32.8% and the Fscore was increased by 22%， which effectively focused on tourists concerns and provided decisionmaking basis for homestay managers and tourism consumers by analyzing the internal causes of positive and negative reviews.

Key words: emotional dictionary; SOPMI; homestay; sentiment analysis; Yangshuo

摘要： 互联网带动旅游相关产业的发展，民宿也因其特色受到游客青睐。分析在线评论中的情感极性，挖掘其潜在价值，可以帮助管理者更好地提高游客的民宿体验。以桂林阳朔的民宿评论为研究对象，使用八爪鱼爬虫软件，爬取携程、去哪儿旅游网站的数据，并以大连理工情感词汇本体库为基础情感词典，加入筛选出的民宿领域情感词，构建民宿领域情感词典。对比基础情感词典，民宿领域情感词典在正向评论方面的精确率提高2.7%，F值提高7.4%；在负面评论方面的精确率提高32.8%，F值提高22%，有效聚焦了游客的关注点。通过分析积极、消极评论的内在原因，为民宿管理者和旅游消费者决策提供依据。