Merge remote-tracking branch 'refs/remotes/honnibal/master'

examples/_handler.py

@@ -0,0 +1,37 @@
+# encoding: utf8
+from __future__ import unicode_literals, print_function
+
+from math import sqrt
+from numpy import dot
+from numpy.linalg import norm
+
+
+def handle_tweet(spacy, tweet_data, query):
+    text = tweet_data.get('text', u'')
+    # Twython returns either bytes or unicode, depending on tweet.
+    # ಠ_ಠ #APIshaming
+    try:
+        match_tweet(spacy, text, query)
+    except TypeError:
+        match_tweet(spacy, text.decode('utf8'), query)
+
+
+def match_tweet(spacy, text, query):
+    def get_vector(word):
+        return spacy.vocab[word].repvec
+
+    tweet = spacy(text)
+    tweet = [w.repvec for w in tweet if w.is_alpha and w.lower_ != query]
+    if tweet:
+        accept = map(get_vector, 'child classroom teach'.split())
+        reject = map(get_vector, 'mouth hands giveaway'.split())
+        
+        y = sum(max(cos(w1, w2), 0) for w1 in tweet for w2 in accept)
+        n = sum(max(cos(w1, w2), 0) for w1 in tweet for w2 in reject)
+        
+        if (y / (y + n)) >= 0.5 or True:
+            print(text)
+
+
+def cos(v1, v2):
+    return dot(v1, v2) / (norm(v1) * norm(v2))

spacy/tokens/token.pyx

@@ -248,24 +248,16 @@ cdef class Token:
 
     property conjuncts:
         def __get__(self):
-            """Get a list of conjoined words"""
+            """Get a list of conjoined words."""
             cdef Token word
-            conjs = []
-            if self.c.pos != CONJ and self.c.pos != PUNCT:
-                seen_conj = False
-                for word in reversed(list(self.lefts)):
-                    if word.c.pos == CONJ:
-                        seen_conj = True
-                    elif seen_conj and word.c.pos == self.c.pos:
-                        conjs.append(word)
-            conjs.reverse()
-            conjs.append(self)
-            if seen_conj:
-                return conjs
-            elif self is not self.head and self in self.head.conjuncts:
-                return self.head.conjuncts
-            else:
-                return []
+            conjuncts = []
+            if self.dep_ != 'conj':
+                for word in self.rights:
+                    if word.dep_ == 'conj':
+                        yield word
+                        yield from word.conjuncts
+                        conjuncts.append(word)
+                        conjuncts.extend(word.conjuncts)
 
     property ent_type:
         def __get__(self):

tests/parser/test_space_attachment.py

@@ -7,6 +7,6 @@ def test_space_attachment(EN):
     sentence = 'This is a test.\nTo ensure  spaces are attached well.'
     doc = EN(sentence)
 
-    for word in doc:
-        if word.is_space:
-            assert word.head.i == (word.i - 1)
+    for sent in doc.sents:
+        if len(sent) == 1:
+            assert not sent[-1].is_space