Understanding Multilingual Correlation of Geo-Tagged Tweets for POI Recommendation

This paper presents a multilingual analysis of Twitter for recommending POIs based on psychographic preferences. People who belong to different countries have different behavioral activities and speak different languages. According to psychographic analysis, for example, people who visit other countries are interested in eating the food of their home country. For this, we aim to clarify psychographic preferences for user behaviors by analyzing geo-tagged tweets based on times, locations, and languages. In this work, we focused on the differences between locations and languages in geo-tagged tweets from European countries. A key feature of the proposed system is the ability to suggest POIs (for tourists) in regions where very few geo-tagged tweets are available in a specific language by using the weighted similarity by others’ preferences. Specifically, we first extract languages of tweets, and we identify tweeting countries based on the latitude and longitude of tweet locations. Then, we extract feature words from tweets of a specific language in a specific region by using a tf-idf based approach. In this paper, we discuss the POI preferences of different language users based on the linguistic correlation between feature words of tweets in the region.