5 ms·

Does a small startup have a chance on hard problems? AtD vs. The Spell Checker Poem

10 points by raffi 17y ago

gregk 17y ago
I tried the demo at http://www.polishmywriting.com/ http://www.polishmywriting.com/ and it found some mistakes I missed in my writing. I am impressed with it's suggestions although it suggested many technical words as spelling errors. Check it out on your writing.
alain94040 17y ago
Sounds like a great OEM play: license your technology so it becomes widespread.
dangoldin 17y ago
It seems in this case you can just break them down into sets of phonemes and do a frequency analysis to get the most common usages. That would take care of the "Eye have"
raffi 17y ago
This is work in progress. I'm working on grammar rules now. Last night though I posted the spell checker poem in just to see how the system would respond to it and I was really surprised (sometimes we're our own harshest critics). I have a plugin for Wordpress too at http://www.afterthedeadline.com http://www.afterthedeadline.com
- jcl 17y ago
  Did it really get the correct word for the first "weigh"? (which is "away", not "way")
  - raffi 17y ago
    Hah, ok score -1 me. It didn't. It came up with way (which was prefixed by a). I think you'll forgive me as that piece is something of a landmine.
lacker 17y ago
This is pretty neat to see. But this is probably the wrong metric to use to measure performance of a spell checker. In the real world you don't want to spell check writing that is 0.5 spelling mistakes per word, and many of these misspellings are not representative of how people actually misspell words. (Who would actually type "mist ache" for "mistake"?) Also, this may just reflect the fact that you have your thresholds for tagging a misspelling set lower. For good metrics you would want to have a large corpus of real documents and measure both the false positive and false negative rates. (To be fair, the author probably knows this stuff already, and is just having fun.)
- raffi 17y ago
  I'm the author: There are two types of spelling errors, real word and non-real word errors. A non-real word error is what most spellcheckers look for. They check if a word is in a dictionary and if it isn't they generate a set of potential words from that error and then rack and stack them. AtD's non-real word spelling corrector uses a dictionary gleaned from the intersection of many wordlists and a corpus of text. The rating of each correction is handled by neural networks. It's all fun and I have carefully measured, tweaked, and remeasured this part of the technology. I also reproduced an experiment conducted by some Polish researchers in 2005 and used their data to find out how I compare to Office 2003. I learned my dictionary needs more work but my accuracy is on par with Office here. The real word errors are just as serious but somehow ignored by most spelling correctors. A real word error is where you use a word that is in the dictionary but really meant some other word. You can imagine why this is difficult. I use a statistical model to try to pick the best word from pre-made confusion sets. While I was impressed with the accuracy of this I had to bias heavily against false positives making it less than what I'd like to see. I'm currently working to have the grammar checker take over or augment some words where this technology does poorly. That said, this poem is all about the real word detection and correction and I was quite delighted to see how well my code did against it anyways (even after I bias heavily against false positives). So in short, yeah I'm just having fun.
hwijaya 17y ago
Just fyi, there are other startups in Australia that also work on this spelling problem http://spellr.us/ http://spellr.us/
- raffi 17y ago
  Judging by the results it looks like they are using Hunspell or Aspell to do the actual checking. I applaud their well executed effort. AtD is more of a complete writing improvement solution that someone like spellr.us could use to provide their service.

10 points by raffi 17y ago

gregk 17y ago

I tried the demo at http://www.polishmywriting.com/ http://www.polishmywriting.com/ and it found some mistakes I missed in my writing. I am impressed with it's suggestions although it suggested many technical words as spelling errors. Check it out on your writing.

alain94040 17y ago

Sounds like a great OEM play: license your technology so it becomes widespread.

dangoldin 17y ago

It seems in this case you can just break them down into sets of phonemes and do a frequency analysis to get the most common usages. That would take care of the "Eye have"

raffi 17y ago

This is work in progress. I'm working on grammar rules now. Last night though I posted the spell checker poem in just to see how the system would respond to it and I was really surprised (sometimes we're our own harshest critics). I have a plugin for Wordpress too at http://www.afterthedeadline.com http://www.afterthedeadline.com

jcl 17y ago
Did it really get the correct word for the first "weigh"? (which is "away", not "way")
- raffi 17y ago
  Hah, ok score -1 me. It didn't. It came up with way (which was prefixed by a). I think you'll forgive me as that piece is something of a landmine.

lacker 17y ago

This is pretty neat to see. But this is probably the wrong metric to use to measure performance of a spell checker. In the real world you don't want to spell check writing that is 0.5 spelling mistakes per word, and many of these misspellings are not representative of how people actually misspell words. (Who would actually type "mist ache" for "mistake"?) Also, this may just reflect the fact that you have your thresholds for tagging a misspelling set lower. For good metrics you would want to have a large corpus of real documents and measure both the false positive and false negative rates. (To be fair, the author probably knows this stuff already, and is just having fun.)

raffi 17y ago
I'm the author: There are two types of spelling errors, real word and non-real word errors. A non-real word error is what most spellcheckers look for. They check if a word is in a dictionary and if it isn't they generate a set of potential words from that error and then rack and stack them. AtD's non-real word spelling corrector uses a dictionary gleaned from the intersection of many wordlists and a corpus of text. The rating of each correction is handled by neural networks. It's all fun and I have carefully measured, tweaked, and remeasured this part of the technology. I also reproduced an experiment conducted by some Polish researchers in 2005 and used their data to find out how I compare to Office 2003. I learned my dictionary needs more work but my accuracy is on par with Office here. The real word errors are just as serious but somehow ignored by most spelling correctors. A real word error is where you use a word that is in the dictionary but really meant some other word. You can imagine why this is difficult. I use a statistical model to try to pick the best word from pre-made confusion sets. While I was impressed with the accuracy of this I had to bias heavily against false positives making it less than what I'd like to see. I'm currently working to have the grammar checker take over or augment some words where this technology does poorly. That said, this poem is all about the real word detection and correction and I was quite delighted to see how well my code did against it anyways (even after I bias heavily against false positives). So in short, yeah I'm just having fun.

hwijaya 17y ago

Just fyi, there are other startups in Australia that also work on this spelling problem http://spellr.us/ http://spellr.us/

raffi 17y ago
Judging by the results it looks like they are using Hunspell or Aspell to do the actual checking. I applaud their well executed effort. AtD is more of a complete writing improvement solution that someone like spellr.us could use to provide their service.