Classification and Visualization of Travel Blog Entries Based on Types of Tourism

We propose a method for classifying travel blog entries into one or more tourism types among six predetermined types by using textual and image information in each entry. Together with this information, we use Wikipedia entries, which are automatically linked from each travel blog entry by entity-linking technology, because information beneficial for classifying blog entries is often mentioned in Wikipedia entries, and we combine this information by using a deep-learning-based method. We conducted an experiment with a neural network using three types of input data. Using the Sparse Composite Document Vector (SCDV) technique, we obtained precision, recall, and F-measure scores of 0.743, 0.217, and 0.336, respectively. We also conducted ensemble learning by using SCDV and support vector machines (SVM), and obtained precision, recall, and F-measure scores of 0.807, 0.179, and 0.293, respectively. Finally, we constructed a system that enables travelers to look for travel blog entries from a map in terms of tourism type.