etBlogAnalysis - Analysing tourism Weblogs and forums using statistical and computer linguistic methods for quality control

Blogs and forums related to travel and tourism have become quite popular in the last years. Travellers use them to share their experiences with other travellers and the whole Web community. Content produced in this way can give an invaluable support in developing and improving tourism products. The number of travel blogs, forums and entries has exploded in recent years. This leads to the uncomfortable situation for tourism managers and marketers that they cannot read and check all the entries potentially mentioning their enterprise, organisation or destination. Semiautomatic routines – software applications, Web crawlers – can be of great help to check the Web automatically based on some criteria defined by the user and create summary reports, which further on can be used to react to entries or use them for marketing purposes. This paper describes the approach used in the etBlogAnalysis project, which aims for developing a combined crawler / sentiment extraction application for the tourism domain. It combines a simple and robust linguistic parsing methodology with information and terminology extraction methods in order to determine relevant utterances on expression (statement) level. The expressions are modelled based on their polarity and their power. The resulting valuated expressions are aggregated on various levels (segment, paragraph, document) and overall reports are produced thereof. A simple application produces warnings for an enterprise (e.g. hotel) if too many negative entries have been written about it.