Investigating the effectiveness of computer-produced summaries obtained from multiple travel blog entries

The evolution of information and communication technology now makes it possible to collect travel information in a variety of ways. Social media content that includes blogs is one such useful information source when planning a trip. This paper proposes a method for generating a summary of multiple travel blog entries that contain images. The proposed method identifies significant sentences in addition to the images by using a graph-based approach that considers travelers’ types of behavior. To investigate the effectiveness of the proposed method, we conducted two experiments: (1) evaluation of the quality of summaries and (2) evaluation of the effectiveness of these summaries. The experimental results demonstrated that the proposed method, LexRank + Image, can outperform state-of-the-art baseline methods for both evaluations. It was also confirmed that the proposed method could produce summaries containing new information not included in travel guidebooks. A system was implemented for generating summaries based on the proposed method.