Boosting Country Classification for Semantic Annotation in Social Networks: Person and Place Country Recognition

摘要

Much research has been done on named entity recognition such as whether the name is a person, company or place, and valuable contributions have been made. However, there has been little research on country recognition of people's names and places. In this paper, we develop a classification technique for social multimedia to automatically classify countries for person or place. This technique will be used in location search, recommendation services, advertisements and country evaluations. Based on binary vector space model (VSM) and boosting algorithm ideas, GBBoosting classification algorithm is designed to support country classification. Since the names for different country multimedia content are very similar sometimes, we construct a weak learner to solve this problem. Compared to weighted similarity and Nai?ve Bayes classification algorithm, GBBoosting classification algorithm is more efficient and has higher recognition rate. GBBoosting classification algorithm has outstanding performance, especially in distinguishing countries with similar spelling.