Ejemplo n.º 1
0
        public static KeyBondMDL <T, L> UpdateKeyBondColl <T, L>(T head, L tail, KeyBondColl <T, L> objKeyBondColl, OffsetWeightMDL objHeadWeightMDL, OffsetWeightMDL objTailWeightMDL)
        {
            if (!objKeyBondColl.Contains(head))
            {
                KeyBondMDL <T, L> bond = new KeyBondMDL <T, L>();
                bond.KeyItem.Key                      = head;
                bond.KeyItem.UpdateOffset             = objKeyBondColl.Parameter.TotalOffset;
                bond.LinkColl.Parameter.ContainerSize = objKeyBondColl.Parameter.ContainerSize;
                bond.LinkColl.Parameter.Threshold     = objKeyBondColl.Parameter.Threshold;
                objKeyBondColl.Add(bond);
            }

            KeyItemMDL <T> mdl = objKeyBondColl[head].KeyItem;

            mdl.ValidCount   = objHeadWeightMDL.Weight + mdl.ValidCount * MemoryDAL.CalcRemeberValue(objKeyBondColl.Parameter.TotalOffset - mdl.UpdateOffset, objKeyBondColl.Parameter);
            mdl.TotalCount   = objHeadWeightMDL.Weight + mdl.TotalCount * MemoryDAL.CalcRemeberValue(1, objKeyBondColl.Parameter);
            mdl.UpdateOffset = objKeyBondColl.Parameter.TotalOffset;


            KeyItemColl <L> objLinkColl = objKeyBondColl[head].LinkColl;

            if (objTailWeightMDL.Offset < 0)
            {
                //继承主计数
                objLinkColl.Parameter.TotalOffset = objKeyBondColl.Parameter.TotalOffset;
                objTailWeightMDL.Offset           = 0;
            }
            KeyItemDAL.UpdateKeyItemColl(tail, objLinkColl, objTailWeightMDL);

            objKeyBondColl.Parameter.TotalValidCount = objHeadWeightMDL.Weight + objKeyBondColl.Parameter.TotalValidCount * MemoryDAL.CalcRemeberValue(objHeadWeightMDL.Offset, objKeyBondColl.Parameter);
            objKeyBondColl.Parameter.TotalOffset    += objHeadWeightMDL.Offset;

            return(objKeyBondColl[head]);
        }
Ejemplo n.º 2
0
        /// <summary>
        /// 使用互信息计算关联系数
        /// </summary>
        /// <typeparam name="T"></typeparam>
        /// <param name="keyHead"></param>
        /// <param name="keyTail"></param>
        /// <param name="objKeyBondColl"></param>
        /// <returns></returns>
        public static double CalcBondRelateValueByPMI <T>(T keyHead, T keyTail, KeyBondColl <T, T> objKeyBondColl)
        {
            ////分别获得相邻单项的频次
            //double dHeadValidCount = objKeyBondColl.Contains(keyHead) ? 1 + objKeyBondColl[keyHead].KeyItem.ValidCount * CalcRemeberValue<T,T>(keyHead,objKeyBondColl) : 1;
            //double dTailValidCount = objKeyBondColl.Contains(keyTail) ? 1 + objKeyBondColl[keyTail].KeyItem.ValidCount * CalcRemeberValue<T,T>(keyTail,objKeyBondColl) : 1;
            //double dTotalValidCount = 1 + 1.0 / ( 1 - MemoryDAL.CalcRemeberValue(1,objKeyBondColl.Parameter) );
            ////获得相邻项共现的频次
            //KeyItemColl<T>  objLinkColl = objKeyBondColl.Contains(keyHead) ? objKeyBondColl[keyHead].LinkColl : new KeyItemColl<T>();

            //KeyItemMDL<T> mdl = objLinkColl.Contains(keyTail) ? objLinkColl[keyTail] : new KeyItemMDL<T>();
            //double dShareValidCount = 1 + mdl.ValidCount * KeyItemHelper.CalcRemeberValue(mdl,objLinkColl);
            //double dShareTotalCount = objKeyBondColl.Contains(keyHead) ? 1 + objKeyBondColl[keyHead].KeyItem.TotalCount : 1;

            //if (!objKeyBondColl.Contains(keyHead) || !objKeyBondColl.Contains(keyTail)) return 0;
            //if (!objKeyBondColl[keyHead].LinkColl.Contains(keyTail)) return 0;

            if (!objKeyBondColl.Contains(keyHead) || !objKeyBondColl.Contains(keyTail))
            {
                return(0);
            }
            if (!objKeyBondColl[keyHead].LinkColl.Contains(keyTail))
            {
                return(0);
            }

            //分别获得相邻单项的频次
            double dHeadValidCount = objKeyBondColl[keyHead].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyHead, objKeyBondColl);
            double dTailValidCount = objKeyBondColl[keyTail].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyTail, objKeyBondColl);

            if (dHeadValidCount < objKeyBondColl.Parameter.Threshold || dTailValidCount < objKeyBondColl.Parameter.Threshold)
            {
                return(0);
            }
            double dTotalValidCount = objKeyBondColl.Parameter.TotalValidCount;// 1.0 / (1 - MemoryDAL.CalcRemeberValue(1, objKeyBondColl.Parameter));

            //获得相邻项共现的频次
            KeyItemColl <T> objLinkColl = objKeyBondColl[keyHead].LinkColl;

            KeyItemMDL <T> mdl = objLinkColl[keyTail];
            double         dShareValidCount = mdl.ValidCount * KeyItemHelper.CalcRemeberValue(mdl, objLinkColl);
            double         dShareTotalCount = objLinkColl.Parameter.TotalValidCount;

            if (dShareTotalCount < objLinkColl.Parameter.Threshold || dShareValidCount < objLinkColl.Parameter.Threshold)
            {
                return(0);
            }

            //P(AB)=P(B|A)*P(A)
            //result=P(AB)/(P(A)*P(B))=P(B|A)/P(B)
            return((dShareValidCount / dShareTotalCount) / (dTailValidCount / dTotalValidCount));
        }
Ejemplo n.º 3
0
        public static double CalcBondRelateValueWithLaplace <T>(T keyHead, T keyTail, KeyBondColl <T, T> objKeyBondColl)
        {
            //分别获得相邻单项的频次
            double dHeadValidCount  = objKeyBondColl.Contains(keyHead) ? 1 + objKeyBondColl[keyHead].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyHead, objKeyBondColl) : 1;
            double dTailValidCount  = objKeyBondColl.Contains(keyTail) ? 1 + objKeyBondColl[keyTail].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyTail, objKeyBondColl) : 1;
            double dTotalValidCount = 1 + 1.0 / (1 - MemoryDAL.CalcRemeberValue(1, objKeyBondColl.Parameter));
            //获得相邻项共现的频次
            KeyItemColl <T> objLinkColl = objKeyBondColl.Contains(keyHead) ? objKeyBondColl[keyHead].LinkColl : new KeyItemColl <T>();

            KeyItemMDL <T> mdl = objLinkColl.Contains(keyTail) ? objLinkColl[keyTail] : new KeyItemMDL <T>();
            double         dShareValidCount = 1 + mdl.ValidCount * KeyItemHelper.CalcRemeberValue(mdl, objLinkColl);
            double         dShareTotalCount = objKeyBondColl.Contains(keyHead) ? 1 + objKeyBondColl[keyHead].KeyItem.TotalCount : 1;


            //P(AB)=P(B|A)*P(A)
            //result=P(AB)/(P(A)*P(B))=P(B|A)/P(B)
            return((dShareValidCount / dShareTotalCount) / (dTailValidCount / dTotalValidCount));
        }
Ejemplo n.º 4
0
        /// <summary>
        /// 当关键项列表发生时,获取关联项一次都不发生时的概率对数,以及包含的关键项
        /// </summary>
        /// <typeparam name="T"></typeparam>
        /// <typeparam name="L"></typeparam>
        /// <param name="objKeyList"></param>
        /// <param name="objMemoryBondColl"></param>
        /// <param name="objLinkProbDict">注:此处存放的是相关项一次都不发生的概率的对数</param>
        /// <param name="objLinkKeyDict"></param>
        public static void UpdateKeyLinkDict <T, L>(List <T> objKeyList, KeyBondColl <T, L> objMemoryBondColl, Dictionary <L, double> objLinkProbDict, Dictionary <L, List <T> > objLinkKeyDict, Dictionary <L, List <int> > objLinkPosDict)
        {
            for (int k = 0; k < objKeyList.Count; k++)
            {
                T key = objKeyList[k];
                if (!objMemoryBondColl.Contains(key))
                {
                    continue;
                }
                KeyItemMDL <T>  objKeyMDL   = objMemoryBondColl[key].KeyItem;
                KeyItemColl <L> objLinkColl = objMemoryBondColl[key].LinkColl;

                double dKeyValidCount = /* objKeyMDL.ValidCount * */ objKeyMDL.TotalCount * CalcRemeberValue <T, L>(objKeyMDL.Key, objMemoryBondColl);

                if (dKeyValidCount < objMemoryBondColl.Parameter.Threshold)
                {
                    continue;
                }
                foreach (KeyItemMDL <L> link in objLinkColl)
                {
                    double dLinkValidCount = link.ValidCount * CalcRemeberValue <L>(link, objLinkColl);
                    if (!objLinkProbDict.ContainsKey(link.Key))
                    {
                        objLinkProbDict.Add(link.Key, 0);
                    }
                    double dLinkProb = dLinkValidCount / dKeyValidCount;
                    objLinkProbDict[link.Key] += dLinkProb >= 1 ? 0 : Math.Log(1 - dLinkProb); //不发生的概率

                    if (!objLinkKeyDict.ContainsKey(link.Key))
                    {
                        objLinkKeyDict.Add(link.Key, new List <T>());
                    }
                    objLinkKeyDict[link.Key].Add(objKeyMDL.Key);

                    if (!objLinkPosDict.ContainsKey(link.Key))
                    {
                        objLinkPosDict.Add(link.Key, new List <int>());
                    }
                    objLinkPosDict[link.Key].Add(k);
                }
            }
        }
Ejemplo n.º 5
0
        /// <summary>
        /// 使用平均信息熵计算关联
        /// </summary>
        /// <typeparam name="T"></typeparam>
        /// <param name="keyHead"></param>
        /// <param name="keyTail"></param>
        /// <param name="objKeyBondColl"></param>
        /// <returns></returns>
        public static double CalcBondRelateValueByAverageEntropy <T>(T keyHead, T keyTail, KeyBondColl <T, T> objKeyBondColl)
        {
            if (!objKeyBondColl.Contains(keyHead) || !objKeyBondColl.Contains(keyTail))
            {
                return(0);
            }
            if (!objKeyBondColl[keyHead].LinkColl.Contains(keyTail))
            {
                return(0);
            }



            //分别获得相邻单项的频次
            double dHeadValidCount = objKeyBondColl[keyHead].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyHead, objKeyBondColl);
            double dTailValidCount = objKeyBondColl[keyTail].KeyItem.ValidCount * CalcRemeberValue <T, T>(keyTail, objKeyBondColl);

            if (dHeadValidCount < objKeyBondColl.Parameter.Threshold || dTailValidCount < objKeyBondColl.Parameter.Threshold)
            {
                return(0);
            }

            //获得相邻项共现的频次
            KeyItemColl <T> objLinkColl = objKeyBondColl[keyHead].LinkColl;

            KeyItemMDL <T> mdl = objLinkColl[keyTail];
            double         dShareValidCount = mdl.ValidCount * KeyItemHelper.CalcRemeberValue(mdl, objLinkColl);
            double         dShareTotalCount = objLinkColl.Parameter.TotalValidCount;

            if (dShareTotalCount < objLinkColl.Parameter.Threshold || dShareValidCount < objLinkColl.Parameter.Threshold)
            {
                return(0);
            }

            double dEuler          = 0.5772156649;
            double dKeywordCount   = objKeyBondColl[keyHead].LinkColl.Count;
            double dAverageEntropy = Math.Log(dKeywordCount) + dEuler - 1;

            double dKeywordEntropy = (dKeywordCount * dShareValidCount / dShareTotalCount) * (Math.Log(dShareTotalCount) - Math.Log(dShareValidCount));

            return(dKeywordEntropy - dAverageEntropy);
        }
Ejemplo n.º 6
0
        public static KeyBondColl <T, L> ClearKeyBondColl <T, L>(KeyBondColl <T, L> objKeyBondColl, bool bIsParentTotalOffset = true)
        {
            double dTotalVaildCount = objKeyBondColl.Parameter.TotalValidCount + 1;// 1.0 / ( 1.0 - MemoryDAL.CalcRemeberValue(1,objKeyBondColl.Parameter) );

            KeyBondColl <T, L> objResultColl = new KeyBondColl <T, L>();

            objResultColl.Parameter         = objKeyBondColl.Parameter;
            objResultColl.Parameter.Entropy = 0;

            double dClearValidCount = 0;

            foreach (KeyBondMDL <T, L> bond in objKeyBondColl)
            {
                KeyItemMDL <T> mdl           = bond.KeyItem;
                double         dRemeberValue = mdl.ValidCount * MemoryDAL.CalcRemeberValue(objKeyBondColl.Parameter.TotalOffset - mdl.UpdateOffset, objKeyBondColl.Parameter);
                if (dRemeberValue > objKeyBondColl.Parameter.Threshold)
                {
                    mdl.ValidCount   = dRemeberValue;
                    mdl.UpdateOffset = 0;

                    if (!objResultColl.Contains(mdl.Key))
                    {
                        dClearValidCount += mdl.ValidCount;

                        double dProb = mdl.ValidCount / dTotalVaildCount;
                        objKeyBondColl.Parameter.Entropy += -dProb *Math.Log(dProb);

                        if (bIsParentTotalOffset)
                        {
                            bond.LinkColl.Parameter.TotalOffset = objKeyBondColl.Parameter.TotalOffset;
                        }
                        bond.LinkColl = KeyItemDAL.ClearKeyItemColl <L>(bond.LinkColl);
                        objResultColl.Add(bond);
                    }
                }
            }
            objResultColl.Parameter.TotalOffset     = 0;
            objResultColl.Parameter.TotalValidCount = dClearValidCount;
            return(objResultColl);
        }
Ejemplo n.º 7
0
        /// <summary>
        /// 分词(同时自动维护词典)
        /// </summary>
        /// <param name="text">待分词文本</param>
        /// <param name="objCharBondColl">邻键集合(用于生成词库)</param>
        /// <param name="objKeyWordColl">词库</param>
        /// <param name="maxWordLen">最大词长(建议:细粒度为4、粗粒度为7)</param>
        /// <param name="bUpdateCharBondColl">是否同时更新邻键集合</param>
        /// <param name="bUpdateKeyWordColl">是否同时更新词库</param>
        /// <param name="nRadiusSize">有效键半径</param>
        /// <returns>返回分词结果</returns>
        public static List <string> Segment(string text, KeyBondColl <string, string> objCharBondColl, KeyItemColl <string> objKeyWordColl, int maxWordLen = 7, bool bUpdateCharBondColl = true, bool bUpdateKeyWordColl = true, int nRadiusSize = 7)
        {
            if (String.IsNullOrEmpty(text))
            {
                return(new List <string>());
            }
            if (maxWordLen <= 0)
            {
                maxWordLen = text.Length;
            }

            //总词频
            double dLogTotalCount = Math.Log(objKeyWordColl.Parameter.TotalValidCount + 1);// Math.Log(1.0 / ( 1.0 - MemoryDAL.CalcRemeberValue(1,objKeyWordColl.Parameter) ) );// Math.Log(objKeyWordColl.Sum(x =>x.VaildCount* KeyWordBLL.CalcRemeberValue<string>(x.Key,objKeyWordColl)));//


            Dictionary <int, List <string> > objKeyWordBufferDict = new Dictionary <int, List <string> >();
            Dictionary <int, double>         objKeyWordValueDict  = new Dictionary <int, double>();

            for (int k = 0; k < text.Length; k++)
            {
                List <string> objKeyWordList = new List <string>();
                double        dKeyWordValue  = 0;

                for (int len = 0; len < maxWordLen; len++)
                {
                    int startpos = k - len;
                    if (startpos < 0)
                    {
                        break;
                    }
                    string keyword = text.Substring(startpos, len + 1);
                    if (len > 0 && !objKeyWordColl.Contains(keyword))
                    {
                        continue;
                    }
                    if (len > 0)
                    {
                        if (!objKeyWordColl.Contains(keyword))
                        {
                            continue;
                        }
                        double dValidCount = KeyItemHelper.CalcValidCount(keyword, objKeyWordColl);
                        if (dValidCount < objKeyWordColl.Parameter.Threshold)
                        {
                            continue;
                        }
                        //if (dValidCount < Math.E) continue;//经测试,原始最好
                    }
                    double dTempValue = 0;
                    if (objKeyWordColl.Contains(keyword))
                    {
                        KeyItemMDL <string> mdl = objKeyWordColl[keyword];
                        dTempValue = -(dLogTotalCount - Math.Log(KeyItemHelper.CalcValidCount(keyword, objKeyWordColl)));
                    }
                    if (objKeyWordValueDict.ContainsKey(startpos - 1))
                    {
                        dTempValue += objKeyWordValueDict[startpos - 1];
                        if (dKeyWordValue == 0 || dTempValue > dKeyWordValue)
                        {
                            dKeyWordValue  = dTempValue;
                            objKeyWordList = new List <string>(objKeyWordBufferDict[startpos - 1]);
                            objKeyWordList.Add(keyword);
                        }
                    }
                    else
                    {
                        if (dKeyWordValue == 0 || dTempValue > dKeyWordValue)
                        {
                            dKeyWordValue  = dTempValue;
                            objKeyWordList = new List <string>();
                            objKeyWordList.Add(keyword);
                        }
                    }
                }
                objKeyWordBufferDict.Add(k, objKeyWordList);
                objKeyWordValueDict.Add(k, dKeyWordValue);

                if (k > maxWordLen)
                {
                    objKeyWordBufferDict.Remove(k - maxWordLen - 1);
                    objKeyWordValueDict.Remove(k - maxWordLen - 1);
                }
            }

            if (bUpdateCharBondColl || bUpdateKeyWordColl)
            {
                KeyWordBLL.UpdateKeyWordColl(text, objKeyWordColl, maxWordLen);
            }

            return(objKeyWordBufferDict[text.Length - 1]);
        }
Ejemplo n.º 8
0
 public static double CalcRemeberValue <T>(KeyItemMDL <T> mdl, KeyItemColl <T> objMemoryBondColl)
 {
     return(MemoryDAL.CalcRemeberValue(objMemoryBondColl.Parameter.TotalOffset - mdl.UpdateOffset, objMemoryBondColl.Parameter));
 }