Font Search Across Various Languages Based on Multimodal Learning

Shun Matsumura, Saemi Choi, Kiyoharu Aizawa

Abstract: We propose a framework that enables to search fonts by using keywords even in the case of minor language fonts without keywords. For example, English letters have a number of font types, and their font data are very well coordinated by being associated with several keywords. Therefore, we can efficiently use the keywords as queries. In contrast, fonts of minor languages are not usually associated with keywords. As a result, users experience difficulty in searching fonts for the best ones. In this paper, we present a methodology comprising two steps to search fonts without keywords: 1) multimodal search of English fonts with keywords; and 2) a search for Japanese fonts across the language. An effective search function was realized at each step. Moreover, in the user test, we verified the effectiveness of font search across various languages.