Flickr image tag understanding and recommendation

• Main Author: Chen, Alicia Ying Ying
• Other Authors: Sun Aixin
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Information systems::Information storage and retrieval
• Online Access: http://hdl.handle.net/10356/62805
• Institution: Nanyang Technological University

Image tagging recommendation systems could be used to present potential tag candidates to users to facilitate their tagging process. However, there is still room for improvement for tag recommender systems to recommend tags that may be more applicable to the user. Therefore, the purpose of this project is to understand how the list of recommended tags would be different if the temporal and location factors are taken into account. The system implemented also provides other functions which can help users to understand the different tagging trends based on the two factors. This project uses a dataset of over 40 million public Flickr images and videos to base its recommendations on. Through the use of Java programming, the system is able to provide functions for analyzing Flickr tags, such as the number of photos containing a tag, its frequency based on geolocation and date, and the popularity of tags in a given time or location. The key function is producing a list of recommended tags by being able to choose if the time and location should be factored into the recommendation process. Three test cases were used to observe whether the recommended tags would be affected by the geolocation or date, and the results are also analyzed in this report to judge the effectiveness of this method.