add tsid channel information for complete BS support.

[rec10/rec10-git.git] / rec10 / trunk / src / n_gram.py

diff --git a/rec10/trunk/src/n_gram.py b/rec10/trunk/src/n_gram.py
index e653d03..f24a097 100644 (file)
--- a/rec10/trunk/src/n_gram.py
+++ b/rec10/trunk/src/n_gram.py
@@ -1,40 +1,120 @@
 #!/usr/bin/python
 # coding: UTF-8
 # Rec10 TS Recording Tools
-# Copyright (C) 2009 Yukikaze
+# Copyright (C) 2009-2011 Yukikaze
+
 import zenhan
-def bigram(str1,str2):
-    str1=zenhan.toHankaku(str1)
-    str2=zenhan.toHankaku(str2)
-    str1=str1.replace(" ","")
-    str2=str2.replace(" ","")
+def bigram(str1, str2):
+    str1 = zenhan.toHankaku(str1)
+    str2 = zenhan.toHankaku(str2)
+    str1 = str1.replace(" ", "")
+    str2 = str2.replace(" ", "")
     """
     bigramによる単語の近さを計算します。
     """
-    gram=[]
-    if len(str1)<2:
+    gram = []
+    if len(str1) < 2:
         gram.append(str1)
     else:
-        tmp=str1[0]
+        tmp = str1[0]
         for x in str1[1:]:
-            gram.append(tmp+x)
-            tmp=x
-    point=0
+            gram.append(tmp + x)
+            tmp = x
+    point = 0
     for x in gram:
-        i=find_gram(x, str2)
-        if i>0:
-            i=90+10*i
+        i = findGram(x, str2)
+        if i > 0:
+            i = 90 + 10 * i
         else:
-            i=0
-        point=point+i
+            i = 0
+        point = point + i
     return point
-def find_gram(gram,s):
+def trigram(str1, str2):
+    """
+    trigramによる単語の近さを計算します。
+    単語文字列の長さも考慮にいれます。
+    """
+    str1 = zenhan.toHankaku(str1)
+    str2 = zenhan.toHankaku(str2)
+    str1 = str1.replace(" ", "")
+    str2 = str2.replace(" ", "")
+    gram = []
+    if len(str1) < 3:
+        gram.append(str1)
+    else:
+        tmp1 = str1[0]
+        tmp2 = str1[1]
+        for x in str1[2:]:
+            gram.append(tmp1 + tmp2 + x)
+            tmp1 = tmp2
+            tmp2 = x
+    point = 0
+    count = 0
+    for x in gram:
+        i = findGram(x, str2)
+        if i > 0:
+            i = 90 + 10 * i
+            count=count + 1
+        else:
+            i = 0
+        point = point + i
+    point = point + 20*count*count
+    point = point / len(str1) * 10
+    point = point / len(str2) * 10
+    return point
+def findGram(gram, s):
     """
     s中にあらわれるgramの数を調べる
     """
-    st=s
-    i=0
-    while st.find(gram)!=-1:
-       i=i+1
-       st=st[st.find(gram)+1:]
-    return i
\ No newline at end of file
+    st = s
+    i = 0
+    while st.find(gram) != -1:
+        i = i + 1
+        st = st[st.find(gram) + 1:]
+    return i
+def getNounQuadGram(s):
+    """
+    。
+    """
+    str1 = zenhan.toHankaku(s)
+    gram={}
+    if len(str1) > 4:
+        tmp1 = str1[0]
+        tmp2 = str1[1]
+        tmp3 = str1[2]
+        skipnum=0
+        for x in str1[3:]:
+            if skipnum>0:
+                skipnum=skipnum-1
+            else:
+                tmps=tmp1 + tmp2 + tmp3 + x
+                if tmps.find(" ")<0:
+                    if zenhan.checkCharacterType(tmp1)>0:
+                        if (zenhan.checkCharacterType(tmp1)==zenhan.checkCharacterType(tmp2)):
+                            if (zenhan.checkCharacterType(tmp2)==zenhan.checkCharacterType(tmp3)):
+                                if (zenhan.checkCharacterType(tmp3)==zenhan.checkCharacterType(x)):
+                                    gram[tmps]=gram.get(tmps,0)+1
+                                    skipnum=0
+                                else:
+                                    skipnum=3
+                            else:
+                                if (zenhan.checkCharacterType(tmp3)==zenhan.checkCharacterType(x)):
+                                    skipnum=2
+                                else:
+                                    skipnum=3
+                        else:
+                            if (zenhan.checkCharacterType(tmp2)==zenhan.checkCharacterType(tmp3)):
+                                if (zenhan.checkCharacterType(tmp3)==zenhan.checkCharacterType(x)):
+                                    skipnum=0
+                                else:
+                                    skipnum=3
+                            else:
+                                if (zenhan.checkCharacterType(tmp3)==zenhan.checkCharacterType(x)):
+                                    skipnum=2
+                                else:
+                                    skipnum=3
+            tmp1 = tmp2
+            tmp2 = tmp3
+            tmp3 = x
+    return gram
+""
\ No newline at end of file