15.01.2013 Views

U. Glaeser

U. Glaeser

U. Glaeser

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

• No toggling between edit/control and ink mode. Pen input inside the input method area is treated<br />

as ink to be recognized by the recognizer; pen input outside this area is treated as mouse events,<br />

e.g., selection, etc. Without this separation, special provisions, sometimes unnatural, have to be<br />

taken to distinguish among the two pen modes.<br />

• Better user control. Within the specially designated writing window it is possible to have additional<br />

GUI elements that help the user with the input task. For instance, there might be buttons<br />

for common edit keys such as backspace, newline, and delete. Similarly, a list of recognition<br />

alternates can be easily displayed and selected from. This is particularly important because top-n<br />

recognition accuracy—a measure of how often the correct answer is among the highest ranked<br />

n results, is generally much higher than top-1 accuracy.<br />

• Consistent UI metaphor. Despite its ergonomic limitations, an on-screen keyboard is generally<br />

available as one of the text input methods on the device. Using a special input area for handwriting<br />

makes the user interface of alternative text entry methods similar.<br />

Modal Input. The possibilities of the user’s input are selectively limited in order to increase recognition<br />

accuracy. Common modes include “digits,” “symbols,” “upper-case letters,” and “lower-case letters”<br />

in English, or “traditional” versus “simplified” in Chinese. By limiting the number of characters<br />

against which a given input ink is matched, the opportunities for confusion and mis-recognition<br />

are decreased, and recognition accuracy is improved. Writing modes represent another tradeoff<br />

between making life simpler for the system or simpler for the user.<br />

Natural Character Set. It is possible to use any character writing style commonly used in the given<br />

language, no need to learn a special alphabet. Characters can be multi-stroke,<br />

i.e., written with<br />

more than one pen trace.<br />

Multi-boxed Input. When multi-stoke input is allowed, end of writing is generally detected by use of<br />

a timer that is set after each stroke is completed; the input is deemed concluded if a set amount<br />

of time elapses before any more input is received in the writing area. This “timeout” scheme is<br />

sometimes confusing to users. Multiple boxes give better performance because a character in one<br />

box can be concluded if input is received in another, removing the need to wait for the timer to<br />

finish.<br />

Of all the restrictions imposed on users by these character-based input methods, modality is the one<br />

where user feedback has been strongest: people want modeless input. The challenge lies in that distinguishing<br />

between letters which have very similar forms across modes is virtually impossible without<br />

additional information. In English orthography, for instance, there are letters for which the lower case<br />

version of the character is merely a smaller version of the upper case version; examples include “Cc,”<br />

“Kk,” “Mm,” “Oo,” “Ss,” “Uu,” “Ww,” etc. Simple attempts at building modeless character recognizers can<br />

result in a disconcerting user experience because upper case letters, or digits, might appear inserted into<br />

the middle of lower case words. Such m1Xed CaSe w0rdS (mixed case words) look to users to be gibberish.<br />

In usability studies, the authors have further found that as the text data entry needs on wireless PIA<br />

devices shifts from short address book or calendar items to longer notes or e-mail messages, users deem<br />

writing one letter at a time to be inconvenient and unnatural.<br />

More Natural User Interfaces<br />

One known way of dealing with the character confusion difficulties described in section “User Interfaces<br />

on Mobile Devices” is to use contextual information in the recognition process. At the simplest level this<br />

means recognizing characters in the context of their surrounding characters and taking advantage of<br />

visual clues derived from word shape. At a higher level, contextual knowledge can be in the form of<br />

lexical constraints, e.g., a dictionary of known words in the language is used to restrict interpretations<br />

of the input ink. These ideas naturally lead to the notion of a word-based text input method. By “word”<br />

we mean a string of characters which, if printed in text using normal conventions, would be surrounded<br />

by white-space characters (see Fig. 42.38(a)).<br />

© 2002 by CRC Press LLC

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!