Take the 2-minute tour ×
English Language & Usage Stack Exchange is a question and answer site for linguists, etymologists, and serious English language enthusiasts. It's 100% free, no registration required.

Claim: a word ending in -y is most likely not a noun but an adjective.

Don't have my tagged corpus handy to check. Anyone have the stats on Parts-Of-Speech of words ending in y and assuming they don't end in -ly?

share|improve this question
add comment

2 Answers

I originally made a quick python script on the "Part of Speech Database" here, which is a combination WordNet and Moby. Then I modified it to run on the frequency list here, based on COCA.

The first script found 29476 words ending in -y, of which 13677 were -ly. Therefore we are left with 15799 words ending in -y but not -ly. Among these words, only 2643 were adjectives.

Therefore our key result is 2643/15799 = 0.16729. Approximately 1 out of 6.

This did not incorporate word frequencies, and I suspected they would boost the ranking somewhat, as many of the -y nonadjectives were quite rare (for example otolaryngology, noun). Thus I edited the program to tally instances of each word from a COCA-derived frequency list.

This found:

  • 23,771,109 instances of -y words;

  • 5,713,230 instances of -ly words;

  • 18,057,879 instances of -y words that were not -ly words;

  • 1,632,165 instances of adjectives among this set.

This leads to a frequency of 1632165/18057879 = 0.090385. Roughly 9% of words ending in -y but not -ly were adjectives. Surprisingly, this result was even smaller. I guess in the scheme of things "traditionally-suffixed" adjectives aren't really that common.

From the data I also found the converse question (does being an adjective generally imply a -y ending?). There were 28426173 total instances of adjectives and 2134139 adjectives ending in -y, including -ly. The result here was quite similar: 0.075077. Only about 3 out of every 40 adjectives have the "traditional" suffix.

share|improve this answer
3  
+1 for the stats. –  Jon Purdy Sep 3 '11 at 20:49
1  
ah, yes, ology. I should have excluded that one as well, but no matter. ly is more common and fairly transparent & regular. I was surprised about how you applied frequency though. As I was reading your answer, I thought you were going to exclude from your calculation words below a certain frequency. Even better would be to calculate the ratio of adj/non-adj for -y words in batches of 1,000. It's well known that POS ratios change as frequency decreases (nouns increase in dominance). I'd guess that -y adjectives tend to comprise a higher ratio at higher frequencies. –  user12549 Sep 3 '11 at 23:23
2  
I'll try the batch thing eventually. For now I have actual work to do. However I do have this: i.imgur.com/rQWzD.png That is a chart of the ratio of instances of adjectives among -y words, excluding words with frequencies below the given cutoff (the bottom axis is in thousands of entries). The data is pretty confounded by the fact that the top 3 adjectives are all imperfect examples (only, early, military). The next five are much better (likely, easy, ready, happy, necessary) but by necessary you've already dipped below 50k. Trying to answer this question quickly may be foolish... –  tdhsmith Sep 4 '11 at 0:52
    
Yes, it is foolish to try to answer it precisely this way, but you've done a commendable job for the level of sophistication you employed. Cheers! –  user12549 Sep 4 '11 at 4:29
2  
@Kyle I admit the stats are arbitrary, but I include them as evidence of the calculation and to show that the final judgment wasn't a close call. The decision to ignore -ly was part of the question so I didn't analyze it too much. Nonetheless, its exclusion seems reasonable to me because I feel it is so much more productive than the other suffixes, and someone/something using the "-y is (not) an adjective" rule would be likely to know about "-ly is an adverb," but not necessarily "-ology is a noun" etc. Also I was thinking about text, so phonological form hadn't even crossed my mind. :P –  tdhsmith Sep 7 '11 at 7:45
show 2 more comments

Frequency results (percent) using WRI curated data.

               ----------------------------------------
                           Word Ending 
               ---------------------------------------
                "y"           "ly"     "y" but not "ly"
Noun           61.58%        17.03%         81.09%
Adverb         24.24%        77.57%          0.88% 
Verb            4.35%         1.06%          5.78% 
Adjective      12.90%         6.46%         15.72%
Interjection    0.40%         0.13%          0.53% 
Determiner      0.12%                        0.17% 
Pronoun         0.06%         0.02%          0.08% 
Preposition     0.02%                        0.03%
Conjunction     0.03%         0.05%

The columns add up more than 100% because the same word can be accounted for in several rows.

enter image description here

Just as a reference, I used the following scripts (only one shown, Mathematica code):

n = Length@Flatten@WordData[___ ~~ "ly", "Lookup"]
{#[[1]], N@#[[2]]/n} & /@ 
  Tally@Flatten@(WordData[#, "PartsOfSpeech"] & /@ 
      WordData[___ ~~ "ly", "Lookup"]) // TableForm
share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.