Multilingual Text Entry using Automatic Language Detection

Multilingual Text Entry using Automatic Language Detection Yo Ehara and Kumiko Tanaka-Ishii Graduate School of Information Science and Technology, University of Tokyo 13F Akihabara Daibiru, 1-18-13 SotoKanda Chiyoda-ku, Tokyo, Japan ehara@r.dl.itc.u-tokyo.ac.jp kumiko@i.u-tokyo.ac.jp Abstract Computer users increasingly need to pro- duce text written in multiple languages. However, typical computer interfaces re- quire the user to change the text entry soft- ware each time a different language is used. This is cumbersome, especially when lan- guage changes are frequent. To solve this problem, we propose TypeAny, a novel front-end interface that detects the language of the user’s key entry and au- tomatically dispatches the input to the ap- propriate text entry system. Unlike previ- ously reported methods, TypeAny can han- dle more than two languages, and can easily support any new language even if the avail- able corpus is small. When evaluating this method, we obtained language detection accuracy of 96.7% when an appropriate language had to be chosen from among three languages. The number of control actions needed to switch languages was decreased over 93% when using Ty- peAny rather than a conventional method. 1 Introduction Globalization has increased the need to produce multilingual text — i.e., text written in more than one language — for many users. When producing a text in a language other than English, a user has to use text entry software corresponding to the other language which will transform the user’s key stroke sequences into text of the desired language. Such software is usually called an input method engine (IME) and is available for each widely used lan- guage. When producing a multilingual text on a typical computer interface, though, the user has to switch IMEs every time the language changes in a multilingual text. The control actions to choose an appropriate IME are cumbersome, especially when the language changes frequently within the text. To solve this problem, we propose a front-end in- terface called TypeAny. This interface detects the language that the user is using to enter text and dy- namically switches IMEs. Our system is situated between the user key entry and various IMEs. Ty- peAny largely frees the user from the need to exe- cute control actions when switching languages. The production of multilingual text involves three kinds of key entry action: • actions to enter text • actions to control an IME1 • actions to switch IMEs Regarding the first and second types, substantial work has been done in the UI and NLP domain, as summarized in (MacKenzie and Tanaka-Ishii, 2007). There has especially been much work re- garding Chinese and Japanese because in these lan- guages the number of actions of the second type is closely related to the accuracy of conversion from Romanized transcription to characters in each of these languages, and this directly reflects the capa- bility of the language model used. 1When using predictive methods such as completion, or kana-kanji conversion in Japanese, the user has to indicate to the IME when it should predict and which proposed …

Content truncated to 500 words