Word completion
From Wikipedia, the free encyclopedia
Word completion is a common feature in web browsers and similar text entry contexts. When a user begins the entry of a frequently-used word, the computer automatically completes it, or proposes a list of choices.
Contents |
[edit] Successful example of word completion
For an example of a relatively successful application of word completion, consider entering "www.microsoft.com" in the address box of a browser. A user may find that typing "www.mi" is sufficient. The completions database in use in this case is the user's browser history. "www.microsoft.com" would be a commonly offered completion. However, other completions may be available, with potential privacy implications, as this list may reveal the browsing history of other users of the computer. Similar problems may occur when word completion is used in predictive text systems as in the example identified by user:MeNext in a predictive text system ITAP marketed by Motorola, which uses word completion.
[edit] Unsuccessful example of word completion
Word completion is not effective for use in predictive text systems. Word completion works well only if there are a small number of possible items to search through. As discussed above, such is the case when entering urls in a browser. But for unrestricted text entry, word completion is an impediment. A word-completion systems requires that the user, after each keystroke, calculates whether it is better to keep on typing to decrease the size of the word-completion list, or examine the best choices presented to try to find their word. This mental effort is often ignored since it is much easier to think simply about counting keystrokes. However the effect of mental computation is real, and causes throughput to decrease rather than increase. Some early work in this area was done by Dunlop and Crossan [1].
Casual thinking about the efficiency of word completion also ignores the effect of errors. Importantly, errors are expensive and common. If a user accepts the wrong word completion, that word will need to be erased in part. Even a few errors can rapidly obliterate any time savings, and tend to be the source of great irritation.
Even if an optimal user existed, who knew for every word and for every keystroke whether it is better to keep typing or examine and select from the list of word completions suggested by the system, they would save at most one keystroke per word on average.
Consider for example entry of the word "soccer" with a word completion system, coupled to a reasonably large dictionary of English. This gives the following result:
| Type | ||
| s | 14,234 words begin with s, | |
| so | 848 words begin with so, | |
| soc | 75 words begin with soc, | |
| socc | 2 words begin with socc |
A user might reasonably guess that stopping at "soc" and then looking though a list of possible completions would find the word "soccer" more quickly than typing the rest of the letters, "cer". Unfortunately, this guess ignores many common words such as "sock" "socket" "society" "social", as well as uncommon words such as socorro or socrates. Such a user would have spend much more time, and become much more frustrated, than they would have had simply typed the rest of the word. Perhaps if they had instead guessed that typing "socc" would be efficient, they might have saved time. But even in that case, they would has spent time and mental effort formulating a guess and plan of action, after every keystroke, time and effort which would have been perhaps better used in composing the message itself.
