Javascript, 1626 bytes:
I wanted to go for a solution which for each character has a clue of which one might come after. Not that short, but no regex and a fairly good result (words: 101 mistakes, non-words, 228 mistakes)
v=function(w){var g={a:{b:3,c:4,d:4,m:6,f:1,r:14,g:4,i:6,x:2,k:2,l:10,n:12,s:6,p:4,t:7,u:2,v:3,w:3,y:3,h:1,z:1},b:{b:3,a:19,e:19,y:3,l:6,o:17,u:12,i:9,s:9,r:6},e:{d:7,l:8,t:4,s:10,n:11,e:10,r:10,c:2,x:2,w:4,a:13,f:1,p:2,g:2,v:1,b:1,m:3,u:1,i:1,k:1,y:2},l:{e:16,y:5,a:16,b:1,f:2,l:12,m:2,o:14,p:1,s:2,u:8,d:4,i:10,k:3,t:5},o:{s:7,g:3,e:3,k:3,n:10,m:4,p:5,w:6,b:3,c:2,t:6,a:5,d:5,h:1,i:2,l:8,o:9,r:8,u:4,y:2,v:2,z:1,f:2,x:1},u:{t:8,e:5,m:7,s:11,a:2,n:13,r:15,d:6,c:4,f:1,g:5,l:9,y:1,z:1,b:5,j:1,x:1,p:2,k:1,i:2},c:{e:9,h:12,i:2,r:6,t:3,o:20,k:15,a:16,l:6,u:8,y:1},h:{e:21,r:2,a:22,i:15,o:20,u:15,n:3,l:1,y:1},i:{d:8,m:5,n:18,r:7,a:2,s:8,v:2,l:13,t:10,b:1,e:6,k:2,p:5,g:3,c:6,o:2,f:2,z:1},m:{e:19,s:8,a:21,i:12,m:1,o:15,y:2,b:4,p:8,n:1,u:8},n:{e:18,u:3,a:9,d:10,n:4,o:7,s:11,t:11,g:10,k:6,i:5,y:2,c:1},r:{e:18,s:4,y:4,a:16,c:1,g:1,i:12,m:3,p:2,t:4,b:1,d:4,k:4,n:5,r:2,o:11,l:2,u:6,f:1},t:{a:14,s:17,e:18,i:9,o:15,h:10,t:3,y:2,c:1,z:1,u:5,r:5,w:2},d:{a:14,d:4,s:10,e:22,y:8,i:12,o:14,r:4,u:10,l:1},f:{a:16,f:6,e:12,y:1,i:14,l:13,o:16,r:7,u:7,t:7,n:1,s:1},g:{a:16,e:12,o:17,u:7,s:18,g:1,y:2,i:8,l:4,n:2,h:3,r:9,w:1},j:{a:25,i:7,e:14,o:25,u:29},k:{i:23,s:6,e:41,u:6,a:10,l:2,n:8,o:2,h:1,y:1},p:{s:12,e:20,a:19,y:2,i:13,t:2,u:10,l:5,o:13,r:4},s:{o:8,i:8,e:13,k:6,h:10,s:8,t:14,y:1,p:5,c:2,l:6,a:10,m:1,n:2,u:4,w:2},v:{a:18,e:47,i:22,o:8,y:6},y:{l:4,e:18,s:20,d:3,n:8,r:8,t:4,a:14,k:1,m:1,o:8,x:3,p:3,u:4,v:1},q:{u:100},w:{a:24,e:17,l:4,r:3,s:10,n:6,y:2,k:1,d:1,t:1,i:17,u:1,o:10,h:4},x:{e:35,i:18,l:6,o:6,a:6,t:12,y:18},z:{z:10,y:10,a:3,e:43,r:3,o:17,i:10,u:3}},p=1,x,y,i=0;for(;i<3;){x=w[i],y=w[++i];p*=g[x]&&g[x][y]||0}return p>60}
Here's a working implementation http://fiddle.jshell.net/jc73sjyn/
In short: the Object g holds the characters from a to z (as keys), and for each of them, there is a set of characters (also as keys) that each represent a character that may come after, along with its probability percentage. Where no object exists, there is no probability.
The 3 scores (4 letters -> 3 evaluations) are multiplied, and a word with a score of 60 and above is considered a real word.
Example: for the word 'cope' there are three lookups:
g[c][o] = 20
g[o][p] = 5
g[p][e] = 20
score = 20 * 5 * 20 = 2000, which is more than 60, so that one is valid.
(I'm quite new with javascript, so there might be ways to make it shorter that I don't know of.)
LATE EDIT:
Completely irrelevant by now, but I evaluated my way to a more correct g:
g={a:{b:7,c:4,d:4,m:6,f:2,r:14,g:4,i:6,x:2,k:2,l:10,n:12,s:6,p:4,t:7,u:2,v:3,w:12,y:3,h:1,z:1},b:{b:10,a:19,e:19,y:3,l:6,o:17,u:10,i:9,s:9,r:3},e:{d:7,l:8,t:4,s:10,n:11,e:10,r:10,c:2,x:2,w:4,a:13,f:1,p:2,g:2,v:20,b:3,m:3,u:1,i:1,k:1,y:2},l:{e:16,y:5,a:16,b:1,f:2,l:12,m:2,o:14,p:1,s:6,u:61,d:1,i:10,k:3,t:5},o:{s:7,g:3,e:3,k:3,n:20,m:4,p:5,w:6,b:3,c:2,t:6,a:5,d:5,h:10,i:2,l:8,o:3,r:8,u:4,y:2,v:2,z:1,f:20,x:1},u:{t:8,e:5,m:7,s:11,a:2,n:13,r:15,d:6,c:1,f:10,g:5,l:9,y:1,z:1,b:5,j:1,x:1,p:2,k:1,i:2},c:{e:9,h:20,i:2,r:6,t:20,o:15,k:15,a:15,l:6,u:8,y:1},h:{e:21,r:2,a:7,i:15,o:20,u:15,n:10,l:0,y:1},i:{d:8,m:5,n:18,r:7,a:5,s:8,v:2,l:13,t:20,b:1,e:21,k:2,p:5,g:20,c:4,o:2,f:2,z:1},m:{e:10,s:8,a:21,i:12,m:1,o:15,y:2,b:4,p:2,n:1,u:8},n:{e:18,u:3,a:9,d:3,n:4,o:20,s:2,t:11,g:10,k:6,i:5,y:2,c:1},r:{e:15,s:4,y:4,a:16,c:1,g:1,i:12,m:3,p:2,t:4,b:1,d:4,k:4,n:5,r:2,o:11,l:2,u:20,f:1},t:{a:14,s:15,e:18,i:2,o:15,h:10,t:3,y:2,c:1,z:1,u:5,r:5,w:2},d:{a:14,d:4,s:10,e:61,y:8,i:12,o:7,r:3,u:10,l:0},f:{a:5,f:6,e:12,y:1,i:3,l:13,o:16,r:7,u:20,t:4,n:1,s:1},g:{a:16,e:12,o:17,u:7,s:18,g:0,y:2,i:8,l:3,n:2,h:3,r:9,w:1},j:{a:8,i:7,e:14,o:5,u:29},k:{i:3,s:20,e:41,u:6,a:10,l:20,n:8,o:2,h:1,y:1},p:{s:12,e:20,a:5,y:2,i:13,t:4,u:10,l:3,o:13,r:4},s:{o:8,i:8,e:13,k:6,h:10,s:8,t:14,y:1,p:5,c:2,l:2,a:10,m:2,n:6,u:8,w:2},v:{a:10,e:20,i:22,o:6,y:6},y:{l:6,e:15,s:20,d:3,n:8,r:8,t:4,a:4,k:1,m:1,o:3,x:3,p:3,u:1,v:1},q:{u:100},w:{a:24,e:17,l:4,r:2,s:3,n:6,y:20,k:1,d:1,t:1,i:17,u:6,o:10,h:20},x:{e:35,i:6,l:3,o:6,a:6,t:3,y:7},z:{z:10,y:10,a:3,e:43,r:1,o:8,i:7,u:1}}
New results:
words: 53 mistakes, non-words: 159 mistakes
http://fiddle.jshell.net/jc73sjyn/2/
... so If I just read your words file into an array and a for-each match on input and break on true, that would satisfy your requirements (and be 100% correct for provided inputs)? Or were you looking for an algorithm? – Bryan Devaney – 2015-02-03T11:21:04.950
4That array has to be part of the code. (I guess). @randomra - care mentioning that you cannot read/access the word list from anywhere.. – Optimizer – 2015-02-03T11:24:35.213
Edited to clarify. – randomra – 2015-02-03T11:30:49.227
3@BryanDevaney you should only hard code 85% of the list – dwana – 2015-02-03T11:54:18.623
Would using the clipboard be "using external resources or reading a file"? – Teun Pronk – 2015-02-03T12:21:40.547
1@TeunPronk Yes. – randomra – 2015-02-03T12:29:33.993
@randomra ah well. It was worth a try since it's not in the loopholes xD – Teun Pronk – 2015-02-03T12:43:48.563
@dwana Just a FYI for those implementing. 85% is a-s. you can safely disgard all inputs starting t-z. – Bryan Devaney – 2015-02-03T13:11:59.167
It's so easy even a monkey could do it
– Mooing Duck – 2015-02-04T22:32:11.473A few answers so far have demonstrated that this is basically probabilistic regex golf. Does it really make sense to keep this as code golf with full program source instead of a regex? – Sparr – 2015-02-05T06:02:57.270
2
@Sparr There's still other ways to do the problem, like in this question
– Sp3000 – 2015-02-05T13:49:26.5975"Is this even a word?" Yes. [<- 4 chars] – chucksmash – 2015-02-05T21:52:22.790
"Is this even a word?" All
even
s are words, therefore the italic one is a word. – ericw31415 – 2016-05-08T22:14:42.257