Evaluating the FOKS Error Model

Slaven Bilac, Timothy Baldwin, Hozumi Tanaka

Abstract: Learners of Japanese face great difficulty when trying to lookup words containing kanji in a dictionary, due to the requirement of knowing the correct reading of the target word. We propose a system that imitates the cognitive process learners go through in generating readings for novel kanji strings, and provide direct access to the dictionary entries based on the generated readings. In doing so we remove the correct reading requirement. The system described here is implemented in a web-based environment and freely available for general use. In this paper we provide an analysis of query and error data collected by our server.