[compare] update formula

hailiang-wang · hailiang-wang · commit c1ac5a3d9c68 · 2018-03-05T14:54:18.000+08:00
diff --git a/VALUATION.md b/VALUATION.md
@@ -1,25 +1,25 @@
 # synonyms 分数评测 [(v3.3.5)](https://pypi.python.org/pypi/synonyms/3.3.5)
 | 词1 |  词2 |   synonyms  |  人工评定 |
 | --- | --- | --- | --- |
-| 轿车 | 汽车 | 0.999  |  0.98 |
-| 宝石 | 宝物 | 0.908  |  0.96 |
-| 旅游 | 游历 | 0.907  |  0.96 |
-| 男孩子 | 小伙子 | 0.999  |  0.94 |
-| 海岸 | 海滨 | 0.909  |  0.925 |
+| 轿车 | 汽车 | 1.0  |  0.98 |
+| 宝石 | 宝物 | 0.888  |  0.96 |
+| 旅游 | 游历 | 0.872  |  0.96 |
+| 男孩子 | 小伙子 | 1.0  |  0.94 |
+| 海岸 | 海滨 | 0.89  |  0.925 |
 | 庇护所 | 精神病院 | 0.511  |  0.9025 |
-| 魔术师 | 巫师 | 0.967  |  0.875 |
-| 中午 | 正午 | 0.91  |  0.855 |
-| 火炉 | 炉灶 | 0.999  |  0.7775 |
-| 食物 | 水果 | 0.8  |  0.77 |
+| 魔术师 | 巫师 | 0.945  |  0.875 |
+| 中午 | 正午 | 0.899  |  0.855 |
+| 火炉 | 炉灶 | 1.0  |  0.7775 |
+| 食物 | 水果 | 0.805  |  0.77 |
 | 鸟 | 公鸡 | 0.699  |  0.7625 |
 | 鸟 | 鹤 | 0.618  |  0.7425 |
-| 工具 | 器械 | 0.908  |  0.7375 |
+| 工具 | 器械 | 0.883  |  0.7375 |
 | 兄弟 | 和尚 | 0.241  |  0.705 |
 | 起重机 | 器械 | 0.485  |  0.42 |
 | 小伙子 | 兄弟 | 0.463  |  0.415 |
 | 旅行 | 轿车 | 0.172  |  0.29 |
 | 和尚 | 圣贤 | 0.474  |  0.275 |
-| 墓地 | 林地 | 0.907  |  0.2375 |
+| 墓地 | 林地 | 0.872  |  0.2375 |
 | 食物 | 公鸡 | 0.441  |  0.2225 |
 | 海岸 | 丘陵 | 0.585  |  0.2175 |
 | 森林 | 墓地 | 0.29  |  0.21 |
diff --git a/demo.py b/demo.py
@@ -92,7 +92,6 @@ def test_similarity(self):
         print("旗帜引领方向 vs 旗帜指引道路:", r)
         # assert r > 0, "the similarity should be bigger then zero"
 
-
         sen1 = "发生历史性变革"
         sen2 = "发生历史性变革"
         r = synonyms.compare(sen1, sen2, seg=True)
diff --git a/setup.py b/setup.py
@@ -13,7 +13,7 @@
 
 setup(
     name='synonyms',
-    version='3.3.5',
+    version='3.3.6',
     description='Chinese Synonyms for Natural Language Processing and Understanding',
     long_description=LONGDOC,
     author='Hai Liang Wang, Hu Ying Xi',
diff --git a/synonyms/__init__.py b/synonyms/__init__.py
@@ -20,7 +20,7 @@
 __copyright__ = "Copyright (c) 2017 . All Rights Reserved"
 __author__ = "Hu Ying Xi<>, Hai Liang Wang<hailiang.hl.wang@gmail.com>"
 __date__ = "2017-09-27"
-__version__ = "3.3.5"
+__version__ = "3.3.6"
 
 import os
 import sys
@@ -68,7 +68,7 @@
 lambda fns
 '''
 # combine similarity scores
-_similarity_smooth = lambda x, y, z: (x * y) + z
+_similarity_smooth = lambda x, y, z, u: (x * y) + z - u
 _flat_sum_array = lambda x: np.sum(x, axis=0)  # 分子
 
 '''
@@ -228,14 +228,18 @@ def _similarity_distance(s1, s2):
     g = cosine(_flat_sum_array(_get_wv(s1)), _flat_sum_array(_get_wv(s2)))
     u = _nearby_levenshtein_distance(s1, s2)
     # print("g: %s, u: %s" % (g, u))
-    if u > 0.8:
-        r = _similarity_smooth(g, 0.005, u)
+    if u >= 0.99:
+        r = 1.0
+    elif u > 0.9:
+        r = _similarity_smooth(g, 0.05, u, 0.05)
+    elif u > 0.8:
+        r = _similarity_smooth(g, 0.1, u, 0.2)
     elif u > 0.4:
-        r = _similarity_smooth(g, 0.05, u)
+        r = _similarity_smooth(g, 0.25, u, 0.15)
     elif u > 0.2:
-        r = _similarity_smooth(g, 0.5, u)
+        r = _similarity_smooth(g, 0.5, u, 0.1)
     else:
-        r = _similarity_smooth(g, 1, u)
+        r = g
 
     if r < 0: r = abs(r)
     r = min(r, 1.0)
@@ -272,6 +276,7 @@ def compare(s1, s2, seg=True):
     seg : True : The original sentences need jieba.cut
           Flase : The original sentences have been cut.
     '''
+    if s1 == s2: return 1.0
     if seg:
         s1 = [x for x in jieba.cut(s1)]
         s2 = [x for x in jieba.cut(s2)]
