add _countweigt

Segment.cpp

@@ -129,7 +129,8 @@ using namespace CppJieba;
 int main()
 {
 	Segment segment;
-	segment.init("dict.utf8");
+	segment.init("jieba.dict.utf8");
+	
 	vector<string> res;
 	string title = "我来到北京清华大学3D电视";
 	bool flag = segment.cutMM(title, res);

Trie.cpp

@@ -12,9 +12,8 @@ namespace CppJieba
         return Trie::iterator(NULL);
     }
 
-    Trie::Trie()
+    Trie::Trie():_root(NULL), _nodeInfoVec(), _totalWeight(0)
     {
-        _root = NULL;
     }
     
     Trie::~Trie()
@@ -22,7 +21,25 @@ namespace CppJieba
         destroy();
     }
 
-    bool Trie::init(const char* const filepath)
+	bool Trie::init(const char* const filePath)
+	{
+		bool res = false;
+		res = _buildTree(filePath);
+		if(!res)
+		{
+			LogError("_buildTree failed.");
+			return false;
+		}
+		res = _countWeight();
+		if(!res)
+		{
+			LogError("_countWeight failed.");
+			return false;
+		}
+		return true;
+	}
+    
+    bool Trie::_buildTree(const char* const filePath)
     {
         char msgBuf[bufSize];
         if(NULL != _root)
@@ -31,7 +48,7 @@ namespace CppJieba
             return false;
         }
         _root = new TrieNode;
-        ifstream ifile(filepath);
+        ifstream ifile(filePath);
         string line;
         vector<string> vecBuf;
         while(getline(ifile, line))
@@ -281,6 +298,31 @@ namespace CppJieba
 		}
         return true;
     }
+
+	bool Trie::_countWeight()
+	{
+		if(_nodeInfoVec.empty() || 0 != _totalWeight)
+		{
+			LogError("_nodeInfoVec is empty or _totalWeight has been counted already.");
+			return false;
+		}
+		
+		//count total freq
+		for(size_t i = 0; i < _nodeInfoVec.size(); i++)
+		{
+			_totalWeight += _nodeInfoVec[i].count;
+			//cout<<_nodeInfoVec[i].word<<_nodeInfoVec[i].count<<endl;
+		}
+		
+		//normalize
+		for(size_t i = 0; i < _nodeInfoVec.size(); i++)
+		{
+			_nodeInfoVec[i].weight = log(double(_nodeInfoVec[i].count)/double(_totalWeight));
+			//cout<<_nodeInfoVec[i].weight<<endl;
+		}
+		
+		return true;
+	}
 }
 
 #ifdef TRIE_UT
@@ -288,9 +330,8 @@ using namespace CppJieba;
 int main()
 {
     Trie trie;
-    trie.init("dict.utf8");
-    char utf[1024] = "我来到北京清华大学3D电视";
-    vector< vector<size_t> > res;
+    trie.init("jieba.dict.utf8");
+    //char utf[1024] = "我来到北京清华大学3D电视";
     trie.destroy();
     return 0;
 }

Trie.h

@@ -5,6 +5,8 @@
 //#include <ext/hash_map>
 #include <map>
 #include <cstring>
+#include <stdint.h>
+#include <cmath>
 #include "cppcommon/str_functs.h"
 #include "cppcommon/vec_functs.h"
 #include "cppcommon/logger.h"
@@ -25,7 +27,8 @@ namespace CppJieba
 		string word;
 		unsigned int count;
 		string tag;
-		TrieNodeInfo():word(),count(0),tag()
+		double weight;
+		TrieNodeInfo():word(),count(0),tag(),weight(0.0)
 		{
 		}
 	};
@@ -91,6 +94,8 @@ namespace CppJieba
         private:
             TrieNode* _root;
 			vector<TrieNodeInfo> _nodeInfoVec;
+			int64_t _totalWeight;
+
         public:
             typedef TrieNodeIterator iterator;
 
@@ -101,7 +106,7 @@ namespace CppJieba
         public:
             Trie();
             ~Trie();
-            bool init(const char* const filepath = DICT_FILE_PATH);
+            bool init(const char* const filePath);
             bool destroy();
             void display();
 
@@ -117,9 +122,11 @@ namespace CppJieba
 			//bool cutMa
 
         private:
+			bool _buildTree(const char* const filePath);
             bool _destroyNode(TrieNode* node);
             void _display(TrieNode* node, int level);
 			bool _insert(const TrieNodeInfo& nodeInfo);
+			bool _countWeight();
 
 		private:
 			enum {bufSize = 1024};