tailieunhanh - Báo cáo khoa học: "Interactive ASR Error Correction for Touchscreen Devices"

We will demonstrate a novel graphical interface for correcting search errors in the output of a speech recognizer. This interface allows the user to visualize the word lattice by “pulling apart” regions of the hypothesis to reveal a cloud of words simlar to the “tag clouds” popular in many Web applications. This interface is potentially useful for dictation on portable touchscreen devices such as the Nokia N800 and other mobile Internet devices. | Interactive ASR Error Correction for Touchscreen Devices David Huggins-Daines Language Technologies Institute Carnegie Mellon University PittsbUrgh PA 15213 UsA dhuggins@ Alexander I. Rudnicky Language Technologies Institute Carnegie Mellon University PittsbUrgh PA 15213 UsA air@ Abstract We will demonstrate a novel graphical interface for correcting search errors in the output of a speech recognizer. This interface allows the user to visualize the word lattice by pulling apart regions of the hypothesis to reveal a cloud of words simlar to the tag clouds popular in many Web applications. This interface is potentially useful for dictation on portable touchscreen devices such as the Nokia N800 and other mobile Internet devices. 1 Introduction For most people dictating continuous speech is considerably faster than entering text using a keyboard or other manual input device. This is particularly true on mobile devices which typically have no hardware keyboard whatsoever a 12-digit keypad or at best a miniaturized keyboard unsuitable for touch typing. However the effective speed of text input using speech is significantly reduced by the fact that even the best speech recognition systems make errors. After accounting for error correction the effective number of words per minute attainable with speech recognition drops to within the range attainable by an average typist Moore 2004 . Moreover on a mobile phone with predictive text entry it has been shown that isolated word dictation is actually slower than using a 12-digit keypad for typing sMs messages Karpov et al. 2006 . 2 Description It has been shown that multimodal error correction methods are much more effective than using speech alone Lewis 1999 . Mobile devices are increasingly being equipped with touchscreens which lend themselves to gesture-based interaction methods. Therefore we propose an interactive method of visualizing and browsing the word lattice using gestures in order to correct .

crossorigin="anonymous">
Đã phát hiện trình chặn quảng cáo AdBlock
Trang web này phụ thuộc vào doanh thu từ số lần hiển thị quảng cáo để tồn tại. Vui lòng tắt trình chặn quảng cáo của bạn hoặc tạm dừng tính năng chặn quảng cáo cho trang web này.