h a l f b a k e r y
It might be better to just get another gerbil.
idea:
add, search, annotate, link, view, overview, recent, by name, best, random
meta:
news, help, about, links, report a problem
account:
Browse anonymously,
or get an account
and write.
Login
Create account.
|
|
|
When using Microsoft Word, one can, if a word is flagged as misspelt, choose to add it to an internal dictionary of words that have been misdiagnosed as wrong.
However, if one proceeds to use another form of this word, either in plural, genitive or something else, more often than not this is also
flagged as incorrect, necessitating one to add the variant to the dictionary as well.
I propose that, when a word is added, the computer also adds other likely forms of that word. If this proves too faulty, the selection of adding to the dictionary could be enhanced, with different options for "Add noun to dictionary", "Add verb to dictionary" etc..
It need not work flawlessly, but I dare say that not being told off every time I used a new form of whatever word I am using would be very convenient nevertheless.
"Stemming"
http://en.wikipedia.org/wiki/Stemming
I don't think this is used in Word, but various search engines utilise algorithms that operate a bit like this. [zen_tom, Feb 03 2008]
[link]
|
| |
Would it mistakenly let you use "gruntled", "underwhelmed", "chalant", "consolate", "wieldy", "descript", "kempt", "shevelled", "maculate", "flappable", "plussed", "cognito", "communicado", "ruly", "ruthful", "pareil", "concerting", "domitable", "corrigible", "committal", "capacitated", "sipid", "petuous", "promptu"... The list goes on. |
|
| |
"couth"
(though the Scottish-born Mrs AWOL tell me that so-and-so may be un-pejoratively described as "couthy") |
|
| |
I'd rather have to "add to dictionary" than risk a misspelling.(-) I have a hard enough time with words that I accidentally approved that I need removed. |
|
| |
I'm with MisterQED. I suspect the
alumnuss of major universitys will be with
me on this one. Some of these problems
could be circumvented, but only if the
software knowed a lot about the whies and
wherefores of English. |
|
| |
Which does not seem to be much of a priority for Microsoft now, let alone making their embedded dictionary complicated enough to understand concatenations. |
|
| |
What would be nice - and please somebody scream "baked" - is a dictionary that recognised all the major variants of English at the same time, rather than focusing on just one of the US/UK/Aus/Can variations (or any others missing from the list). I've had to home-brew this in the past with a merger. |
|
| |
I think that's called the OED, [boysparks]. |
|
| |
Sorry, should have been clearer. I meant the kind of dictionary used for spell-checking by word processors and browsers. |
|
| |
As an example, my version of MS Word allows 'Language' to be selected as English (US) or English (UK) but not both simultaneously. |
|
| |
It probably sounds daft to want such a feature, but as someone who works with more than one variant, it would be a neat touch to have the facility to accept either spelling, as well as to be able to adjust all spellings to match a particular version; e.g. "Anglicise This". |
|
| |
Although, no doubt if it does exist, that feature ironically reads as "Anglicize This". |
|
| |
I find that selecting English (UK) or English (Aus) still gives me Americanis(z)ed auto-corrections of words I know to be spelled a certain way in the common usage of those countries. It's possibly the single most annoying "feature" of MS Office. |
|
| |
I leave spellchecker turned off. It helps with my blood pressure control. |
|
| |
Back to the idea posted, perhaps the application could allow the user to |
|
| |
(2) Add word stem and extensions |
|
| |
Option (2) would display a suggested stem and its extensions in a column of user-editable fields. Each field would be labelled ("Stem", "Plural", "Past Tense", and so on), with a common word and its extensions displayed in a parallel column to clarify the intent. Unwanted fields could just be blanked out by the user. |
|
| |
Problem is that a word not already in the dictionary is likely to be a proper noun, abbreviation, or slang, and as such is unlikely to 'play by the rules'. So it's probably not any quicker to do this than to just separately spell-check all variations used in the document. |
|
| |
Later: [+] an evolutionary computing approach to this, breeding and selecting algorithms, might produce half-decent results for very little effort. There's lots of training data already out there (i.e. the dictionary) and a clearly defined fitness function. |
|
| |
I think there have been a number of attempts to create neural networks that scan through a set of training documents and determine, through contextual positioning and some amount of stem-awareness, a stochastic view of how well constructed English is formed. |
|
| |
I'm thinking more of a competitive co-evolutionary approach that just focuses on correctly identifying stem and extensions for a given word. |
|
| |
The competitive and co-evolutionary aspect would arise through the evolution of two species: The extension deducing algorithms and the training sets. |
|
| |
The idea is that as the algorithms are selected over time for their ability to correctly deduce extensions for the training sets, so too are the training sets selected for their 'toughness'. Effectively, an arms race. There are a few checks that need to be built into such a model to prevent niching, but it's all very do-able. |
|
| |