N-Best Hidden Markov Model Supertagging to Improve Typing on an Ambiguous Keyboard
Saša Hasan, Karin Harbusch · 2004
Ambiguous keyboards, i.e. several letters host on the same key (cf. telephone keyboard), produce candidate lists with words that match the entered code. These lists have to be disambiguated by the user for the intended word. Consequently, the primary goal is to order these candidates in a way that the most appropriate words are placed on top of the suggestion list for minimal selection costs or to postpone the attention to suggestions to a final editing step. This paper reports on promising results for this goal by inspecting the whole sentence on the basis of supertagging and lightweight dependency analysis.