Add support to load from GloVe models.

docs/Embeddings.rst

@@ -18,6 +18,7 @@ The Embedding class can read word embeddings from different sources:
 
 -  Gensim word2vec objects: (``from_gensim`` method)
 -  Word2vec binary/text models: (``from_word2vec`` method)
+-  GloVe models (``from_glove`` method)
 -  polyglot pickle files: (``load`` method)
 
 .. code:: python

polyglot/mapping/embeddings.py

@@ -247,6 +247,46 @@ def from_word2vec(fname, fvocab=None, binary=False):
 
     return Embedding(vocabulary=vocabulary, vectors=vectors)
 
+  @staticmethod
+  def _from_glove(fname):
+    with _open(fname, 'rb') as fin:
+      words = []
+      vocab_size, layer1_size = None, None
+      vectors = []
+      for line_no, line in enumerate(fin):
+        try:
+          parts = unicode(line, encoding="utf-8").strip().split()
+        except TypeError as e:
+          parts = line.strip().split()
+        except Exception as e:
+          logger.warning("We ignored line number {} because of erros in parsing"
+                          "\n{}".format(line_no, e))
+          continue
+        # We deduce layer1_size because GloVe files have no header.
+        if layer1_size is None:
+          layer1_size = len(parts) - 1
+        # We differ from Gensim implementation.
+        # Our assumption that a difference of one happens because of having a
+        # space in the word.
+        if len(parts) == layer1_size + 1:
+          word, weights = parts[0], list(map(float32, parts[1:]))
+        else:
+          logger.warning("We ignored line number {} because of unrecognized "
+                          "number of columns {}".format(line_no, parts[:-layer1_size]))
+          continue
+        index = line_no
+        words.append(word)
+        vectors.append(weights)
+      vectors = np.asarray(vectors, dtype=np.float32)
+      return words, vectors
+
+
+  @staticmethod
+  def from_glove(fname):
+    words, vectors = Embedding._from_glove(fname)
+    vocabulary = OrderedVocabulary(words)
+    return Embedding(vocabulary=vocabulary, vectors=vectors)
+
   @staticmethod
   def load(fname):
     """Load an embedding dump generated by `save`"""