Proposal of Word Prediction Method for Gaze Swipe Text Entry

Gaze swipe, one of the text entry methods that do not require physical interaction with a device, involves sequentially tracing the characters of the desired word on a displayed keyboard using the user’s gaze. This approach holds the promise of enabling quick word entry. However, previous gaze swipe method presents an issue where the time required for typing increases with longer words, consequently burdening the user. As a solution, there is a demand for an input word prediction method that presents the desired word for typing before the user completes tracing all the characters with their gaze. Thus, this paper proposes a gaze swipe typing method incorporating word prediction functionality. Through experimentation, the base gaze swipe typing method’s typing speed was measured at 8.30 words per minute, with an accuracy rate of 91% for the top 3 estimated words. Additionally, for the proposed method, when 70% of the gaze path length was typed, the accuracy rate for the top 3 predicted words was confirmed to be 63%.