Open Source Language Checking Technology
May 3, 2010 1:23 PM   Subscribe

After The Deadline is an open source spell/style/grammar checker from Automattic for WordPress, Firefox and other stuff.

Here's a demo of the Firefox plugin and here's a (rather poor quality) presentation from WordCamp NYC 2009. I saw this guy's presentation at WordCamp SF and thought it was pretty neat.
posted by brundlefly (28 comments total) 14 users marked this as a favorite
 
I have a few co-workers who are getting this installed on their workstations whether they like it or not.
posted by bizwank at 1:28 PM on May 3, 2010 [1 favorite]


Will it make sure that that I case Improved Magic Missile instead of regular Magic Missile?
posted by oddman at 1:28 PM on May 3, 2010 [1 favorite]


Guess I'm out of a job! Anyone want to rob banks with me?
posted by theredpen at 1:33 PM on May 3, 2010 [4 favorites]


I'm very intrepid in this new spellbinder if it's smarter than the one i coherently utile.
posted by fuq at 1:36 PM on May 3, 2010 [7 favorites]


I thought firefox already checked my spelling? If firefox isn't doing it, who keeps putting those red lines under my typos?
posted by Hoenikker at 1:43 PM on May 3, 2010


I'd install this but it would delulzify my webpostages.
posted by doublesix at 1:47 PM on May 3, 2010


Firefox corrects spelling, but doesn't go after grammar. If I had to guess why, it's because spelling is easier to extend to most languages than grammar.
posted by pwnguin at 1:47 PM on May 3, 2010


who keeps putting those red lines under my typos?

God. He is sick people misspelling "celebrate" as "celibate". That particular mistake has caused no end of misery over the years so He's taking an active interest from now on.
posted by quin at 1:48 PM on May 3, 2010 [1 favorite]


Hoenikker: “If firefox isn't doing it, who keeps putting those red lines under my typos?

Elves. They overplayed their hand in the last round of salary negotiations, though.
posted by Kadin2048 at 1:53 PM on May 3, 2010


Will it make sure that that I case Improved Magic Missile instead of regular Magic Missile?

Perhaps not, but you still might find it useful.
posted by peeedro at 2:02 PM on May 3, 2010


Hoenikker: "who keeps putting those red lines under my typos?"

Terrence Malick
posted by brundlefly at 2:02 PM on May 3, 2010 [2 favorites]


Pretty cool, but it doesn't appear to catch some glaring errors. So the employment of editors is safe... for now. Here's my test paragraph, with noted errors underlined:

Abraham Lincoln, did not want to loose the civil war. He bravely lead his troop's into battle, and they're military might ultimately won the day. The Confederate forces were stretched to thin, and the Unions victory was eminent.

I'm impressed it got "eminent "-> "imminent". It failed to similarly correct immanent, though. "Were stretched" produced a suggestion that I shouldn't use the passive voice.

Once this works better, maybe we can use GreaseMonkey to transparently correct all YouTube comments...
posted by qxntpqbbbqxl at 2:10 PM on May 3, 2010


I gave it a quick test with one of my blog entries -- caught a few things (mostly minor, casual language things) and I thought it was pretty useful. It's definitely not a replacement for human editors, but I like it since it's always hard to read my own stuff and spot problems, especially after I've just written something.

Cool idea.
posted by darksong at 2:16 PM on May 3, 2010


What I find most interesting about the product is that it's not a standalone spell-checker. I was pretty much expecting it to just be a wrapper around the venerable iSpell library, with some sort of grammar checking tacked on. It's not.

What it seems to be, at least on first glance, is a client/server system where one backend does checking for multiple clients. (The easy-install versions aimed at end users take advantage of a backend run by Automattic.) That could be awesome or scary depending on your point of view: on the plus side, you get a constantly-evolving, intelligent dictionary and grammar rules corpus that's fed by many users. On the minus side, you could end up with a fair amount of what you write, or at least key terms / SIPs getting sent to a remote server somewhere. If you're writing anything personal, that could be a little weird. Although lots of people use Gmail for very personal stuff (myself included) and there's no reason to immediately distrust Automattic any more than you distrust Google. I would be interested in getting their take on some of the privacy implications, though.

Several places that I've worked at have had shared Word spelling dictionaries. I'm not sure exactly how it's accomplished, but it was pretty slick to sit down to an otherwise brand-new computer and have it immediately recognize and correct all the institutional acronyms and technical terms that everyone there used. If Automattic's product works like that but on a larger scale, it could be pretty neat.
posted by Kadin2048 at 2:20 PM on May 3, 2010


So, a client/server spellchecker over the internet to be analyzed? No privacy implications there!
posted by delmoi at 2:44 PM on May 3, 2010


Of course it is open source, so you can run your own server, but apparantly you need to build your own corpus in order to get it running. Huh.
posted by delmoi at 2:55 PM on May 3, 2010


My input: Its' after three in the afternoon.
The analysis: No writing errors were found.

~Sigh~ Still no tech solution for the apostrophically challenged.
posted by vapidave at 3:43 PM on May 3, 2010


This bookmarklet does not work with Google Chrome.

Looks like you guys will just have to live with my typos, then.
posted by PhoBWanKenobi at 3:49 PM on May 3, 2010


Very cool. I will take any editing I can get.
posted by bearwife at 3:51 PM on May 3, 2010


Eh, the idea's nice, and something that's both open-source and likely to raise the standard of writing on the Internet should be applauded, but it's let down by the grammar checker being ridiculously poor. It has the same aversion to the passive voice that Word has, for one thing, and when I pasted my Biggles/CSI Miami crossover fanfic into the demonstration it kept telling me I should entirely drop words like "onto", which would have made no sense. Not impressed, basically.
posted by ZsigE at 4:00 PM on May 3, 2010


The difficulty with grammar checking is that the computer simply does not have any understanding whatsoever. Considering that the English language has some fairly illogical rules this is not surprising.

I think they will invent a perfect grammar checker the same day they invent a camera that automatically takes great photographs.
posted by Sukiari at 8:20 PM on May 3, 2010


There is no such thing as a computerized grammar or style checker. There are only programs that randomly underline portions of your text and spit out generic bad advice.
posted by moss at 8:57 PM on May 3, 2010 [1 favorite]


spelling
posted by jpcooper at 12:19 AM on May 4, 2010


I don't think the difficulty of writing a good grammar checker stems from English's illogical rules (a computer is at least as capable of keeping track of a huge list of rules and weird exceptions as a human is). It stems from the fact that a lot of grammatical errors produce sentences that are perfectly grammatical but have an unlikely meaning. It's perfectly valid to write that someone might not want to loose a war (perhaps after crying havoc). Or that their victory is immanent, if you're taking a sort of philosophical attitude towards the war. So you can't fix those errors without an understanding of what the person is writing, and sometimes not even then — even a human editor who's an expert in the topic at hand and comprehends the whole work being written can have trouble with this.

Spelling-checkers, on the other hand, are possible because English text has a huge amount of letter-by-letter redundancy; each letter carries only 2-3 bits of actual information, out of maybe 5 bits to represent the letter. And even spelling-checkers are just barely functional (I mistyped 'letter' as 'latter' back there, for example.)
posted by hattifattener at 12:23 AM on May 4, 2010 [2 favorites]


So you can't fix those errors without an understanding of what the person is writing, and sometimes not even then — even a human editor who's an expert in the topic at hand and comprehends the whole work being written can have trouble with this.

This kind of tool isn't going to work perfectly, or even to be at a state in which it can avoid great criticism, anytime soon. I don't see why the author saw fit to stop at just grammar, though--a lot of the weird problems which this checker has difficulty with (the kinds of problems which hattifattener addressed above) seem to me to be more within the fields of semantics and pragmatics. Some sort of semantic wondertool, perhaps one that underlined your horrible constructions with shades of a color to emphasize subtleties in different meanings, is probably closer to what people expect.
posted by a sourceless light at 1:09 AM on May 4, 2010


...and when you COMBINE the semantics and pragmatics puzzlers with spelling- and grammar-checkers, you could maybe get something pretty close to what five seconds of glancing over your own writing can do!
posted by a sourceless light at 1:12 AM on May 4, 2010


Wordpress, eh? So the resulting text will crash when more than 2 people try to read it and one month later will have sneaky links to spam text inserted in 6pt at the bottom, then.
posted by bonaldi at 8:49 AM on May 4, 2010


Hello, I'm the one who wrote After the Deadline and presented it at WordCamp San Francisco. I see a few points that came up in the conversation that I'd like to discuss:

Privacy:

After the Deadline uses a software as a service architecture. This means your requests are sent to our server for processing. This is necessary because the checks AtD performs are CPU and memory intensive. We don't save your requests. We process them and send your errors and suggestions for them. Automattic has a privacy policy that AtD follows: http://automattic.com/privacy/ And AtD is open source (as someone pointed out), you're welcome to look at the code or run your own server.

AtD vs. The Firefox Spell Checker vs. other spell checkers:

1. Most spell checkers do not look at context when making suggestions. After the Deadline does. This means you're more likely to see the word you meant suggested when you make an error.

2. Many spell checkers have dictionaries that lack many words that are common on the web. AtD's dictionary is made from recent examples of writing that include modern terms like "blog".

3. AtD detects misused homophones by looking at the context you used them in. This is how it found imminent vs. eminent. No other web writing tool has this feature. This is my favorite feature and the one I'm most proud of.

Style checking is optional:

PolishMyWriting.com is a demo of the AtD technology. You're firehosed with everything it can highlight including passive voice and hidden verbs. In WordPress and Firefox, the style checker (blue underlines), hidden verb checker, and passive voice finder are optional and off by default. I believe anyone can benefit from spell and grammar check. Style checking gets personal. If you want it, it's there. If not, it's off. :) I find it valuable for technical writing.

Please, rethink your relationship with your spell checker:

No tool will catch every mistake you make. It's not possible for reasons others have pointed out here. It's another source of feedback on your writing. That said, there is a benefit you may not be aware of. After the Deadline will train you to be a better writer.

When you mistake a homophone and it catches it, an explain button is available. This explanation has the definitions of the words. Most grammar errors have this explain button available too. For example, if you mistake an indefinite article, AtD will explain what an indefinite article is and when to use a or an. This program will fill gaps in your grammar knowledge as you're writing. The style checker will change your inner-voice if you let it.

The goal of this project is not to catch every mistake, it's to train writers to avoid these mistakes in the first place.

I hope you enjoy it.

-- Raphael
posted by rsmudge at 9:16 AM on May 4, 2010 [3 favorites]


« Older Lynn Redgrave Passes Away   |   SEC Prophecy in 2004 Newer »


This thread has been archived and is closed to new comments