[update] Change the behavior of Sapphire class

- Sapphire class returns only 1 word alignment & 1 phrase alignment - Some fixes in *_alignment.py according to changes in Sapphire class
m-yoshinaka · Sep 16, 2020 · 1ea07f6 · 1ea07f6
1 parent b58f304
commit 1ea07f6
Show file tree

Hide file tree

Showing 4 changed files with 31 additions and 38 deletions.
diff --git a/sapphire/phrase_alignment.py b/sapphire/phrase_alignment.py
@@ -158,4 +158,4 @@ def _forward(s, t, start_node, end_node, pairs):
 
         alignments.sort(key=lambda x: float(x[1]), reverse=True)
 
-        return alignments
+        return alignments[0][0]
diff --git a/sapphire/sapphire.py b/sapphire/sapphire.py
@@ -1,36 +1,21 @@
-from . import setting
 from .word_alignment import FastTextVectorize, WordAlign
 from .phrase_alignment import PhraseExtract, PhraseAlign
 
 
-class SapphireAlignment(object):
-
-    def __init__(self, word_alignment: list, top_alignment: list):
-        self.name = ''
-        self.word_alignment = word_alignment
-        self.top_alignment = top_alignment
-
-
 class Sapphire(object):
 
     def __init__(self, model):
-        self.name = ''
-
-        self._hungarian = setting.HUGARIAN
-        self._lambda = setting.LAMBDA
-        self._delta = setting.DELTA
-        self._alpha = setting.ALPHA
-
         self.vectorizer = FastTextVectorize(model)
         self.word_aligner = WordAlign()
         self.extractor = PhraseExtract()
         self.phrase_aligner = PhraseAlign()
+        self.set_params()
 
-    def set_params(self, LAMBDA, DELTA, ALPHA, HUGARIAN=False):
-        self._lambda = LAMBDA
-        self._delta = DELTA
-        self._alpha = ALPHA
-        self._hungarian = HUGARIAN
+    def set_params(self, lambda_=0.6, delta=0.6, alpha=0.01, hungarian=False):
+        self.lambda_ = lambda_
+        self.delta = delta
+        self.alpha = alpha
+        self.hungarian = hungarian
 
     def align(self, tokens_src: list, tokens_trg: list):
         len_src = len(tokens_src)
@@ -39,12 +24,18 @@ def align(self, tokens_src: list, tokens_trg: list):
         vectors_src = self.vectorizer.vectorize(tokens_src)
         vectors_trg = self.vectorizer.vectorize(tokens_trg)
 
-        sim_matrix = self.word_aligner.similarity_matrix(vectors_src, vectors_trg)
-        word_alignment = self.word_aligner.align(sim_matrix, self._lambda, self._hungarian)
-
-        phrase_pairs = self.extractor.extract(word_alignment, vectors_src, vectors_trg, self._delta, self._alpha)
-        phrase_alignment = self.phrase_aligner.create_lattice(phrase_pairs, len_src, len_trg)
-
-        result = SapphireAlignment(word_alignment, phrase_alignment[:10])
-
-        return result
+        sim_matrix = self.word_aligner.similarity_matrix(vectors_src,
+                                                         vectors_trg)
+        word_alignment = self.word_aligner.align(sim_matrix,
+                                                 self.lambda_,
+                                                 self.hungarian)
+
+        phrase_pairs = self.extractor.extract(word_alignment,
+                                              vectors_src,
+                                              vectors_trg,
+                                              self.delta,
+                                              self.alpha)
+        phrase_alignment = self.phrase_aligner.create_lattice(phrase_pairs,
+                                                              len_src,
+                                                              len_trg)
+        return word_alignment, phrase_alignment
diff --git a/sapphire/word_alignment.py b/sapphire/word_alignment.py
@@ -18,16 +18,16 @@ class FastTextVectorize(WordEmbedding):
     def __init__(self, model):
         super().__init__()
         self.model = model
+        self.dim = model.get_dimension()
 
     def vectorize(self, words: list) -> np.array:
         vector = []
 
         if words:
             for word in words:
                 vector.append(self.model.get_word_vector(word.lower()))
-                # vector.append(self.model.get_word_vector(word))
         else:
-            vector.append(np.zeros(300))
+            vector.append(np.zeros(self.dim))
 
         return np.array(vector)
 
@@ -99,7 +99,8 @@ def _final(matrix):
 
         align_matrix = np.logical_and(src2trg, trg2src)
         union_matrix = np.logical_or(src2trg, trg2src)
-        neighbors = [(-1, 0), (0, -1), (1, 0), (0, 1), (-1, -1), (-1, 1), (1, -1), (1, 1)]
+        neighbors = [(-1, 0), (0, -1), (1, 0), (0, 1),
+                     (-1, -1), (-1, 1), (1, -1), (1, 1)]
 
         _grow_diag()
         _final(src2trg)

diff --git a/setup.py b/setup.py
@@ -11,11 +11,12 @@ def read_requirements():
 
 setup(
     name='sapphire',
-    version='0.1.0',
-    description='Simple Aligner for Phrasal Paraphrase with Hierarchical Representation',
+    version='0.1.1',
+    description='Simple Aligner for Phrasal Paraphrase \
+                 with Hierarchical Representation',
     author='Masato Yoshinaka',
     author_email='yoshinaka.masato@ist.osaka-u.ac.jp',
     install_requires=read_requirements(),
-    url='https://github.com/mybon13/sapphire',
+    url='https://github.com/m-yoshinaka/sapphire',
     packages=find_packages()
-)
+)
Original file line number	Diff line number	Diff line change
Expand Up		@@ -158,4 +158,4 @@ def _forward(s, t, start_node, end_node, pairs):

		alignments.sort(key=lambda x: float(x[1]), reverse=True)

		return alignments
		return alignments[0][0]