Tests for the evaluate_word_pairs function

gensim/models/word2vec.py

@@ -1402,7 +1402,7 @@ def log_evaluate_word_pairs(pearson, spearman, oov, pairs):
         return KeyedVectors.log_evaluate_word_pairs(pearson, spearman, oov, pairs)
 
     def evaluate_word_pairs(self, pairs, delimiter='\t', restrict_vocab=300000, case_insensitive=True, dummy4unknown=False):
-        return self.wv.evaluate_word_pairs(self, pairs, delimiter, restrict_vocab, case_insensitive, dummy4unknown)
+        return self.wv.evaluate_word_pairs(pairs, delimiter, restrict_vocab, case_insensitive, dummy4unknown)
 
     def __str__(self):
         return "%s(vocab=%s, size=%s, alpha=%s)" % (self.__class__.__name__, len(self.wv.index2word), self.vector_size, self.alpha)
@@ -1629,4 +1629,3 @@ def __iter__(self):
         model.accuracy(args.accuracy)
 
     logger.info("finished running %s", program)
-

gensim/test/test_data/wordsim353.tsv

@@ -0,0 +1,355 @@
+# The WordSimilarity-353 Test Collection (http://www.cs.technion.ac.il/~gabr/resources/data/wordsim353/)
+# Word 1	Word 2	Human (mean)
+love	sex	6.77
+tiger	cat	7.35
+tiger	tiger	10.00
+book	paper	7.46
+computer	keyboard	7.62
+computer	internet	7.58
+plane	car	5.77
+train	car	6.31
+telephone	communication	7.50
+television	radio	6.77
+media	radio	7.42
+drug	abuse	6.85
+bread	butter	6.19
+cucumber	potato	5.92
+doctor	nurse	7.00
+professor	doctor	6.62
+student	professor	6.81
+smart	student	4.62
+smart	stupid	5.81
+company	stock	7.08
+stock	market	8.08
+stock	phone	1.62
+stock	CD	1.31
+stock	jaguar	0.92
+stock	egg	1.81
+fertility	egg	6.69
+stock	live	3.73
+stock	life	0.92
+book	library	7.46
+bank	money	8.12
+wood	forest	7.73
+money	cash	9.15
+professor	cucumber	0.31
+king	cabbage	0.23
+king	queen	8.58
+king	rook	5.92
+bishop	rabbi	6.69
+Jerusalem	Israel	8.46
+Jerusalem	Palestinian	7.65
+holy	sex	1.62
+fuck	sex	9.44
+Maradona	football	8.62
+football	soccer	9.03
+football	basketball	6.81
+football	tennis	6.63
+tennis	racket	7.56
+Arafat	peace	6.73
+Arafat	terror	7.65
+Arafat	Jackson	2.50
+law	lawyer	8.38
+movie	star	7.38
+movie	popcorn	6.19
+movie	critic	6.73
+movie	theater	7.92
+physics	proton	8.12
+physics	chemistry	7.35
+space	chemistry	4.88
+alcohol	chemistry	5.54
+vodka	gin	8.46
+vodka	brandy	8.13
+drink	car	3.04
+drink	ear	1.31
+drink	mouth	5.96
+drink	eat	6.87
+baby	mother	7.85
+drink	mother	2.65
+car	automobile	8.94
+gem	jewel	8.96
+journey	voyage	9.29
+boy	lad	8.83
+coast	shore	9.10
+asylum	madhouse	8.87
+magician	wizard	9.02
+midday	noon	9.29
+furnace	stove	8.79
+food	fruit	7.52
+bird	cock	7.10
+bird	crane	7.38
+tool	implement	6.46
+brother	monk	6.27
+crane	implement	2.69
+lad	brother	4.46
+journey	car	5.85
+monk	oracle	5.00
+cemetery	woodland	2.08
+food	rooster	4.42
+coast	hill	4.38
+forest	graveyard	1.85
+shore	woodland	3.08
+monk	slave	0.92
+coast	forest	3.15
+lad	wizard	0.92
+chord	smile	0.54
+glass	magician	2.08
+noon	string	0.54
+rooster	voyage	0.62
+money	dollar	8.42
+money	cash	9.08
+money	currency	9.04
+money	wealth	8.27
+money	property	7.57
+money	possession	7.29
+money	bank	8.50
+money	deposit	7.73
+money	withdrawal	6.88
+money	laundering	5.65
+money	operation	3.31
+tiger	jaguar	8.00
+tiger	feline	8.00
+tiger	carnivore	7.08
+tiger	mammal	6.85
+tiger	animal	7.00
+tiger	organism	4.77
+tiger	fauna	5.62
+tiger	zoo	5.87
+psychology	psychiatry	8.08
+psychology	anxiety	7.00
+psychology	fear	6.85
+psychology	depression	7.42
+psychology	clinic	6.58
+psychology	doctor	6.42
+psychology	Freud	8.21
+psychology	mind	7.69
+psychology	health	7.23
+psychology	science	6.71
+psychology	discipline	5.58
+psychology	cognition	7.48
+planet	star	8.45
+planet	constellation	8.06
+planet	moon	8.08
+planet	sun	8.02
+planet	galaxy	8.11
+planet	space	7.92
+planet	astronomer	7.94
+precedent	example	5.85
+precedent	information	3.85
+precedent	cognition	2.81
+precedent	law	6.65
+precedent	collection	2.50
+precedent	group	1.77
+precedent	antecedent	6.04
+cup	coffee	6.58
+cup	tableware	6.85
+cup	article	2.40
+cup	artifact	2.92
+cup	object	3.69
+cup	entity	2.15
+cup	drink	7.25
+cup	food	5.00
+cup	substance	1.92
+cup	liquid	5.90
+jaguar	cat	7.42
+jaguar	car	7.27
+energy	secretary	1.81
+secretary	senate	5.06
+energy	laboratory	5.09
+computer	laboratory	6.78
+weapon	secret	6.06
+FBI	fingerprint	6.94
+FBI	investigation	8.31
+investigation	effort	4.59
+Mars	water	2.94
+Mars	scientist	5.63
+news	report	8.16
+canyon	landscape	7.53
+image	surface	4.56
+discovery	space	6.34
+water	seepage	6.56
+sign	recess	2.38
+Wednesday	news	2.22
+mile	kilometer	8.66
+computer	news	4.47
+territory	surface	5.34
+atmosphere	landscape	3.69
+president	medal	3.00
+war	troops	8.13
+record	number	6.31
+skin	eye	6.22
+Japanese	American	6.50
+theater	history	3.91
+volunteer	motto	2.56
+prejudice	recognition	3.00
+decoration	valor	5.63
+century	year	7.59
+century	nation	3.16
+delay	racism	1.19
+delay	news	3.31
+minister	party	6.63
+peace	plan	4.75
+minority	peace	3.69
+attempt	peace	4.25
+government	crisis	6.56
+deployment	departure	4.25
+deployment	withdrawal	5.88
+energy	crisis	5.94
+announcement	news	7.56
+announcement	effort	2.75
+stroke	hospital	7.03
+disability	death	5.47
+victim	emergency	6.47
+treatment	recovery	7.91
+journal	association	4.97
+doctor	personnel	5.00
+doctor	liability	5.19
+liability	insurance	7.03
+school	center	3.44
+reason	hypertension	2.31
+reason	criterion	5.91
+hundred	percent	7.38
+Harvard	Yale	8.13
+hospital	infrastructure	4.63
+death	row	5.25
+death	inmate	5.03
+lawyer	evidence	6.69
+life	death	7.88
+life	term	4.50
+word	similarity	4.75
+board	recommendation	4.47
+governor	interview	3.25
+OPEC	country	5.63
+peace	atmosphere	3.69
+peace	insurance	2.94
+territory	kilometer	5.28
+travel	activity	5.00
+competition	price	6.44
+consumer	confidence	4.13
+consumer	energy	4.75
+problem	airport	2.38
+car	flight	4.94
+credit	card	8.06
+credit	information	5.31
+hotel	reservation	8.03
+grocery	money	5.94
+registration	arrangement	6.00
+arrangement	accommodation	5.41
+month	hotel	1.81
+type	kind	8.97
+arrival	hotel	6.00
+bed	closet	6.72
+closet	clothes	8.00
+situation	conclusion	4.81
+situation	isolation	3.88
+impartiality	interest	5.16
+direction	combination	2.25
+street	place	6.44
+street	avenue	8.88
+street	block	6.88
+street	children	4.94
+listing	proximity	2.56
+listing	category	6.38
+cell	phone	7.81
+production	hike	1.75
+benchmark	index	4.25
+media	trading	3.88
+media	gain	2.88
+dividend	payment	7.63
+dividend	calculation	6.48
+calculation	computation	8.44
+currency	market	7.50
+OPEC	oil	8.59
+oil	stock	6.34
+announcement	production	3.38
+announcement	warning	6.00
+profit	warning	3.88
+profit	loss	7.63
+dollar	yen	7.78
+dollar	buck	9.22
+dollar	profit	7.38
+dollar	loss	6.09
+computer	software	8.50
+network	hardware	8.31
+phone	equipment	7.13
+equipment	maker	5.91
+luxury	car	6.47
+five	month	3.38
+report	gain	3.63
+investor	earning	7.13
+liquid	water	7.89
+baseball	season	5.97
+game	victory	7.03
+game	team	7.69
+marathon	sprint	7.47
+game	series	6.19
+game	defeat	6.97
+seven	series	3.56
+seafood	sea	7.47
+seafood	food	8.34
+seafood	lobster	8.70
+lobster	food	7.81
+lobster	wine	5.70
+food	preparation	6.22
+video	archive	6.34
+start	year	4.06
+start	match	4.47
+game	round	5.97
+boxing	round	7.61
+championship	tournament	8.36
+fighting	defeating	7.41
+line	insurance	2.69
+day	summer	3.94
+summer	drought	7.16
+summer	nature	5.63
+day	dawn	7.53
+nature	environment	8.31
+environment	ecology	8.81
+nature	man	6.25
+man	woman	8.30
+man	governor	5.25
+murder	manslaughter	8.53
+soap	opera	7.94
+opera	performance	6.88
+life	lesson	5.94
+focus	life	4.06
+production	crew	6.25
+television	film	7.72
+lover	quarrel	6.19
+viewer	serial	2.97
+possibility	girl	1.94
+population	development	3.75
+morality	importance	3.31
+morality	marriage	3.69
+Mexico	Brazil	7.44
+gender	equality	6.41
+change	attitude	5.44
+family	planning	6.25
+opera	industry	2.63
+sugar	approach	0.88
+practice	institution	3.19
+ministry	culture	4.69
+problem	challenge	6.75
+size	prominence	5.31
+country	citizen	7.31
+planet	people	5.75
+development	issue	3.97
+experience	music	3.47
+music	project	3.63
+glass	metal	5.56
+aluminum	metal	7.83
+chance	credibility	3.88
+exhibit	memorabilia	5.31
+concert	virtuoso	6.81
+rock	jazz	7.59
+museum	theater	7.19
+observation	architecture	4.38
+space	world	6.53
+preservation	world	6.19
+admission	ticket	7.69
+shower	thunderstorm	6.31
+shower	flood	6.03
+weather	forecast	8.34
+disaster	area	6.25
+governor	office	6.34
+architecture	century	3.78

gensim/test/test_word2vec.py

@@ -378,6 +378,18 @@ def testAccuracy(self):
         kv_accuracy = model.wv.accuracy(datapath('questions-words.txt'))
         self.assertEqual(w2v_accuracy, kv_accuracy)
 
+    def testEvaluateWordPairs(self):
+        """Test Spearman and Pearson correlation coefficients give sane results on similarity datasets"""
+        corpus = word2vec.LineSentence(datapath('head500.noblanks.cor.bz2'))
+        model = word2vec.Word2Vec(corpus, min_count=3, iter=10)
+        correlation = model.evaluate_word_pairs(datapath('wordsim353.tsv'))
+        pearson = correlation[0][0]
+        spearman = correlation[1][0]
+        oov = correlation[2]
+        self.assertTrue(0.1 < pearson < 1.0)
+        self.assertTrue(0.1 < spearman < 1.0)
+        self.assertTrue(0.0 <= oov < 90.0)
+
     def model_sanity(self, model, train=True):
         """Even tiny models trained on LeeCorpus should pass these sanity checks"""
         # run extra before/after training tests if train=True