
The file "word_relationship.questions" contains the syntactic analogies.
For example,
"good better rough"
with the interpretation "good is to better as rough is to what?"

The file "word_relationship.answers" contains the answers along with 
the specific type of relation tested. 
For example,
"JJ_JJR rougher"
This means the first word was a base adjective, and the second a 
comparative adjective.

Note that the patterns are systematically created, as described in the 
text. For example, the patterns based on good/better/best with the contrast
base "rough" are:

good better rough
better good rougher
good best rough
best good roughest
best better roughest
better best rougher

The file "sample.output" contains sample answers to each of the 8000 questions.
There is one hypothesis per line. To score this, run "sample.scoring_script.sh".
The output should look like:

$ ./sample.scoring_script.sh
JJ_JJR 15.8
JJR_JJ 10.4

JJ_JJS 7.6
JJS_JJ 5

JJR_JJS 5.6
JJS_JJR 10.8

Adjectives 9.2 % correct

NNS_NN 4.4
NN_NNS 7.4

NN_NNPOS 18.2
NNPOS_NN 14.4

Nouns 11.1 % correct

VB_VBZ 16.4
VBZ_VB 18.8

VB_VBD 16.2
VBD_VB 17.8

VBD_VBZ 18.2
VBZ_VBD 17

Verbs 17.4 % correct

Overall correct: 12.75 %

------------------------------------------

Citation:
Thomas Mikolov, Wen-tau Yih and Geoffrey Zweig
Linguistic Regularities in Continuous SpaceWord Representations
In Proceedings, The 2013 Conference of the North American Chapter of the
Association for Computational Linguistics: Human Language Technologies 
(NAACL-HLT 2013)

