Nationality Identification from Names Using N-gram Based Cumulative Frequency Addition

摘要

This paper describes two N-gram based classification methods -Cumulative Frequency Addition and Naive Bayesian - that identify a person's nationality, or at least the nationality group, from his/her name. Language classification using N-gram based methods has been shown to be highly accurate and insensitive to typographical errors, and, as a result, these methods have been extensively researched and documented in the natural language processing literature. However, there has been little research in using names to identify nationality, especially using N-gram based linguistic features. The two classifiers described here accomplish that goal efficiently on name data from the 2004 Olympic Games. Although similar in speed and accuracy, the novel Cumulative Frequency Addition method is somewhat simpler than the more conventional Naive Bayesian method. We obtained accuracies of 86% on a 14 country database and 96% on a 7 country database within the top 3 choices, which we argue is sufficient for applications like Text to Speech systems for significantly improving name pronunciation.