Hacker News new | ask | show | jobs
by stcredzero 5014 days ago
> Figure out a way to route all of your writing through your search engine, and have it flag phrases which do not occur in its database, or which occur much less frequently than similar ones. So, in this case, it would ask: "Did you mean 'desert planet'?"

Or, one could build a table of trigrams and bi-grams for English words and use that in the spell checker. It would be helpful to have some sort of approximate indexing, so you can catch things like 'dessert planet' which is just 1 edit step away from 'desert planet.'
1 comments
ThaddeusQuay2 5014 days ago
My suggestion was for Blekko, where he works. It just occurred to me that there are interesting possibilities in connecting a search engine with a word processor. Anyway, even though your table/index thingy would work, it's availability and/or accuracy might suffer from whatever manual intervention which would be necessary for maintenance, but mostly, it would not have the advantage of constant updates from crawling new and updated sources. There is more to English than the OED. A while back, I saw an estimate of there being something like more than one million words in English, which included many terms that have come into normal usage, but which might take years to be included in any mainstream dictionary.
link
stcredzero 5014 days ago
> There is more to English than the OED. A while back, I saw an estimate of there being something like more than one million words in English...

I can see the advantage of a cloud-hosted solution here. It's getting pretty compute/memory intensive. Still, "spell/grammar" check is something that we'll probably never get completely right. We'll just peck away and approach asymptotically. Desert/dessert and their ilk are a worthy target. (Words that "look" right.)

link
ThaddeusQuay2 5014 days ago
"We'll just peck away and approach asymptotically."

My pigeons like to peck at things, and they seem to have success with that approach. We could probably learn something from them.

link
ChuckMcM 5014 days ago
I caught the reference, and have been idly thinking about ways to plug the generic AJAX style text edit box into an API that allows for spelling analysis. (this is more analysis than correction, as you point out). Not surprisingly if you get aggressive about the corrections folks complain.
link
ThaddeusQuay2 5014 days ago
"folks complain"

I wouldn't, as long as all of the extras can be toggled on/off. I'd actually pay a monthly fee for a search engine which has many features and is very customizable.

link