Raffi's Blog <3

Dear friends,

I’m writing to announce that After the Deadline is officially launched.  We’ve made great strides recently with a grammar checker, new website, and a fixed up TinyMCE plugin.  Thank you for your bug reports and suggestions that have shaped this project.

Now that we’re launched, I’d like to ask you to help spread the word.  If you like AtD, please tell people about it.  These links should help tell the story:

What does After the Deadline do?
http://www.afterthedeadline.com/features.slp

After the Deadline – Media Kit
http://www.afterthedeadline.com/mediakit.slp

As usual, if you have any questions please contact me and I’ll do what I can to help.  Thank you again.

– Raphael

After the Deadline, Dashnine Media

Last week was another busy week for After the Deadline.

I posted an update to the rules on Friday.  This update expanded the coverage of the grammar checker.  New confused words were added.  I added support for flagging double negatives. I also updated the dictionary with several hyphenated words.

This update also means you’ll see more style suggestions.  I lowered the score suggestions must pass to get displayed.  I also cleaned up the rules that forced me to make the score so high to begin with.   After the Deadline started life as a style checker and I intend to make sure you benefit from this feature.

The webpage also saw several updates last week.  I added a features page with demonstrations of what After the Deadline does.  Send it to your friends.  I also fixed a number of issues behind the scenes (the kind I’m embarrassed to admit–like the forgot password feature not working).

I’m working on the FAQ and preparing for launch.  Expect more details tomorrow.

After the Deadline

This week’s update to AtD is quite full featured.  I think you’ll like it.  Here is a brief summary:

• I rewrote the error explanations.  The passive voice and hidden verb explanations have useful information now!
• Several bugs were fixed–AtD now does better processing single quotes and spellchecking possessive forms of words
• Several annoying style checker rules were updated–the style checker is the oldest part of the AtD codebase.  I’m working to make the style checker smarter using the features added for the grammar checker.
• Tons of new grammar rules were added.  These include expanded determiner and auxiliary verb agreement, confused words, and some subject-verb agreement rules.

After the Deadline

There were some troubles with the support@afterthedeadline.com email address [*cough* the account never existed until tonight *cough*].  I just want you to know this has been fixed.  Contact through the After the Deadline website and the toll free phone number were not affected.  If you sent an email to that address consider resending it.  My apologies for this mishap.

After the Deadline

This week was filled with insane progress for After the Deadline.  Besides the new website and the updated plugins there was significant progress on the backend as well.

Language Model

I staged a server devoted to the tasks of recompiling After the Deadline’s language model, training the neural networks, and reporting the system’s accuracy.  I then added ten million words worth of data to the mix and pushed the giant red button.  I’m happy to report that the system responded favorably to the extra data.

Agreement Rules

This week I also worked on auxiliary verb and determiner-noun agreement.

Auxiliary verbs expect a past participle verb form.  This isn’t a real problem as most verbs have the same past form as their past participle form.  Unfortunately there are pesky irregular verbs that have a different past participle form and these are a common source of writing errors.  Here is some of what I found in Wikipedia-Simple:

RAW is WWE’s main show and has ran on Monday nights since 1993
the cultivation of the rose flower is thought to have began in Iran
The houses were build again after the war

The corrections for these sentences are “has run”, “have begun”, and “were built”.

Certain determiners expect a singular noun and others expect a plural noun.  I’ve started working on detecting errors with the wrong kind of determiner/noun combination (they’re called agreement errors for a reason).  Here is a sample of what I’m finding in Wikipedia-Simple:

This pages is more or less a copy of the EnWP article
The differents opinions are called controversies
The problems is the misunderstanding

I plan to continue this work supporting different determiners and finding errors with them next week.  These new rules were pushed to the production server tonight.   Enjoy!

After the Deadline

After the Deadline is rolling forward like a road paver with a jet engine attached.  Today I’ve posted updates to the Wordpress and TinyMCE plugins.

Wordpress Plugin Updates

If you’re using Wordpress you win big.  Not only do you get the TinyMCE updates and a flashier icon, I’ve also added an “Ignore always” menu.  Now when you click on an error you have the option of telling it to stay away.  This information is kept in a cookie stored in your browser.  You can also unignore phrases by visiting the After the Deadline configuration under Plugins -> After the Deadline in the Wordpress admin area.  You will need to download and reinstall the After the Deadline plugin to get this new functionality.

Many of you have requested this and now you have it.  Enjoy!

TinyMCE Plugin Updates

TinyMCE is a WYSIWYG editor used in many web applications (Wordpress included).  In this update I rewrote the code for highlighting phrases reported by After the Deadline.  The new code is cleaner and I assume less buggy.  I briefly tested it against Firefox, Safari, and IE.  Everything seems in order but if you notice problems please report them.

I’ve also added a TinyMCE option atd_ignore_strings.  You can set this value in your TinyMCE init function to tell the editor which phrases it should ignore.  I’ve included code to show how to setup the “Ignore always” menu that I recently added to Wordpress.  It’s up to you if you want to give end users this functionality or not.

See the AtD TinyMCE Plugin README for more information.

After the Deadline

I’ve made crazy progress on After the Deadline this week.

This update features new rules for who’s vs. whose, their vs. there, your vs. you’re, where vs. were, and to vs. too.   These rules performed well flagging the most common misuses of these words in my test corpus.  Here are some examples from Wikipedia-Simple English:

just because your an admin on the
a hyphen is also used when a word is too long too fit in one row of writing
but mentally challenged might be to hard to understand
they where named the orange revolution

AtD now converts plural nouns to their singular form in some suggestions.  What does this mean for you?  I’ve added rules that detect the use of a plural after a and an.   For example, AtD will revise “they aimed a weapons at us” to “they aimed a weapon at us” or “they aimed weapons at us”.

senate has a judges as chairman and four assessors
in protestant churches a saints is anyone who is religious
a regions is a part of the country with a government and a president

Enjoy the updates.

After the Deadline

Have you ever seen the spell checker poem? It’s a funny poem meant to highlight the shortcomings of computer assisted writing technology.  Here it is:

The Spelling Chequer (or poet tree without mist takes)

Eye have a spelling chequer
It came with my pea sea
It plainly marks four my revue
Miss steaks eye cannot see

Each thyme when I have struct the quays
Eye weight four it two say
If watt eye rote is wrong or rite
It shows me strait a weigh

As soon as a mist ache is maid
It nose bee fore too late
And eye  can put the error rite
Eye really fined it grate

I’ve run this poem threw it
I’m sure your policed to no
It’s letter perfect in its weigh
My chequer tolled me sew

- Author Unknown

Oh a whim I decided to run this through Microsoft Office, Firefox, and After the Deadline.  Care to see how well they do?

Firefox vs. The Spell Checker Poem

Firefox does the worst of the three.  The suggestion for the word Chequer is “Chewer” and struct is not flagged to the right word.  This is understandable because Firefox’s spellchecker does not look at the surrounding words when providing suggestions.

Microsoft Word 2004 for Mac vs. The Spell Checker Poem

Here are the results for Microsoft Word.  I build all my technology on an ancient Powerbook so this is the best I have.  Here Microsoft Word 2004 suggests the word Chaucer for Chequer.  And don’t let the fancy green underlining fool you, the grammar checker is useless here too.  The recommendation for “Eye have” is “Eye has”.  I think a mist ache is my favorite as the grammar checker suggests mists ache.   So it is a toss up here who is doing worse Firefox or Microsoft Word.

I’d like to be fair though.  The newest version of Microsoft Word includes a contextual spellchecker that is supposed to catch the kind of errors contained in this poem.

After the Deadline vs. The Spell Checker Poem

Two of the claim to fame features in After the Deadline are contextual spellchecking and real word misuse detection.  I don’t claim perfection in my real word misuse detection but this poem is a good example of how it works.  See for yourself:

All the flagged words actually have the correct word as the first suggestion.  I’m asked all the time why does After the Deadline matter?  We have Microsoft Office and our browser has spell check, right?

And now, for a shameless plug…

After the Deadline is a software service that adds grammar, style, and spell checking to web applications. You can embed it or download a plugin for common applications. A plugin for Wordpress is available at http://www.afterthedeadline.com and you can try it at http://www.polishmywriting.com

After the Deadline

I have good news. AtD now has a grammar checking engine and I am actively adding rules.  Tonight I pushed rules for a vs. an, it’s vs. its, and one case of where vs. were.   This removes a big burden from the real word error detection feature and will allow me to make this feature more accurate.  AtD now has the potential to find a variety of writing errors beyond misused words.

AtD’s grammar checker is an exciting technology.  It is rule based using word and part-of-speech information to find phrases of interest.  It then uses a statistical language model to filter poor suggestions.  This language model allows AtD to automatically infer rule exceptions automatically.

This is a very smart engine and I hope you’re as excited as I am about what is coming next.

After the Deadline

My part-of-speech tagger and I were going to take over the world.   Reality set in when I discovered that the bigram word model used in AtD is better than part-of-speech tags at making predictions about what belongs.

What does this mean for you?  It means I’m working on the grammar checker now.

Yay.

After the Deadline