/// <summary>
///根据孤立点拆分长句，然后再分别对各个句子的片段进行分词.
///长中文句子的分词困扰了我3年，一直没有好的解决方案。没想到在观看
///2010年世界杯开幕式时，我突发灵感，想出了这个孤立点分割拆分长句的
///算法，彻底解决的这个长期困扰我的难题.
///eaglet 11th Jun 2010 注释留念
/// </summary>
/// <param name="positionLenArr">保护位置和长度信息的单词分量数组</param>
/// <param name="orginalTextLength">原始字符串长度</param>
/// <param name="count">positionLenArr 的 count</param>
/// <returns></returns>
private Node[] GetLeafNodeArray(微博舆论.Dict.PositionLength[] positionLenArr, int orginalTextLength, int count)
{
//Split by isolated point
Node[] result = new Node[TopRecord];
int lastRightBoundary = positionLenArr[0].Position + positionLenArr[0].Length;
int lastIndex = 0;
for (int i = 1; i < count; i++)
{
if (positionLenArr[i].Position >= lastRightBoundary)
{
//last is isolated point
int c = i - lastIndex;
微博舆论.Dict.PositionLength[] arr = new 微博舆论.Dict.PositionLength[c];
Array.Copy(positionLenArr, lastIndex, arr, 0, c);
Node[] leafNodeArray = GetLeafNodeArrayCore(arr, lastRightBoundary - positionLenArr[lastIndex].Position, c);
Framework.QuickSort <Node> .TopSort(leafNodeArray, _LeafNodeList.Count, (int)Math.Min(TopRecord, _LeafNodeList.Count), new NodeComparer(_Options.FrequencyFirst));
CombineNodeArr(result, leafNodeArray);
lastIndex = i;
}
int newRightBoundary = positionLenArr[i].Position + positionLenArr[i].Length;
if (newRightBoundary > lastRightBoundary)
{
lastRightBoundary = newRightBoundary;
}
}
if (lastIndex < count)
{
//last is isolated point
int c = count - lastIndex;
微博舆论.Dict.PositionLength[] arr = new 微博舆论.Dict.PositionLength[c];
Array.Copy(positionLenArr, lastIndex, arr, 0, c);
Node[] leafNodeArray = GetLeafNodeArrayCore(arr, lastRightBoundary - positionLenArr[lastIndex].Position, c);
Framework.QuickSort <Node> .TopSort(leafNodeArray, _LeafNodeList.Count, (int)Math.Min(TopRecord, _LeafNodeList.Count), new NodeComparer(_Options.FrequencyFirst));
CombineNodeArr(result, leafNodeArray);
}
return(result);
}
public SuperLinkedList <WordInfo> Match(微博舆论.Dict.PositionLength[] positionLenArr, string orginalText, int count)
{
if (_Options == null)
{
_Options = Setting.PanGuSettings.Config.MatchOptions;
}
if (_Parameters == null)
{
_Parameters = Setting.PanGuSettings.Config.Parameters;
}
int[] masks = new int[orginalText.Length];
int redundancy = _Parameters.Redundancy;
SuperLinkedList <WordInfo> result = new SuperLinkedList <WordInfo>();
if (count == 0)
{
if (_Options.UnknownWordIdentify)
{
WordInfo wi = new WordInfo();
wi.Word = orginalText;
wi.Position = 0;
wi.WordType = WordType.None;
wi.Rank = 1;
result.AddFirst(wi);
return(result);
}
else
{
int position = 0;
foreach (char c in orginalText)
{
WordInfo wi = new WordInfo();
wi.Word = c.ToString();
wi.Position = position++;
wi.WordType = WordType.None;
wi.Rank = 1;
result.AddLast(wi);
}
return(result);
}
}
Node[] leafNodeArray = GetLeafNodeArray(positionLenArr, orginalText.Length, count);
//下面两句是不采用孤立点分割算法的老算法
//Node[] leafNodeArray = GetLeafNodeArrayCore(positionLenArr, orginalText.Length, count);
//Framework.QuickSort<Node>.TopSort(leafNodeArray,
// _LeafNodeList.Count, (int)Math.Min(TopRecord, _LeafNodeList.Count), new NodeComparer());
int j = 0;
// 获取前TopRecord个单词序列
foreach (Node node in leafNodeArray)
{
if (leafNodeArray[j] == null)
{
break;
}
if (j >= TopRecord || j >= leafNodeArray.Length)
{
break;
}
Dict.PositionLength[] comb = new 微博舆论.Dict.PositionLength[node.AboveCount];
int i = node.AboveCount - 1;
Node cur = node;
while (i >= 0)
{
comb[i] = cur.PositionLength;
cur = cur.Parent;
i--;
}
_AllCombinations.Add(comb);
j++;
}
//Force single word
//强制一元分词
if (_Options.ForceSingleWord)
{
Dict.PositionLength[] comb = new 微博舆论.Dict.PositionLength[orginalText.Length];
for (int i = 0; i < comb.Length; i++)
{
微博舆论.Dict.PositionLength pl = new 微博舆论.Dict.PositionLength(i, 1, new WordAttribute(orginalText[i].ToString(), POS.POS_UNK, 0));
pl.Level = 3;
comb[i] = pl;
}
_AllCombinations.Add(comb);
}
if (_AllCombinations.Count > 0)
{
ICollection <Dict.PositionLength> positionCollection = MergeAllCombinations(redundancy);
foreach (Dict.PositionLength pl in positionCollection)
//for (int i = 0; i < _AllCombinations[0].Length; i++)
{
//result.AddLast(new WordInfo(_AllCombinations[0][i], orginalText));
result.AddLast(new WordInfo(pl, orginalText, _Parameters));
if (pl.Length > 1)
{
for (int k = pl.Position;
k < pl.Position + pl.Length; k++)
{
masks[k] = 2;
}
}
else
{
masks[pl.Position] = 1;
}
}
}
#region 合并未登录词
bool needRemoveSingleWord;
List <WordInfo> unknownWords = GetUnknowWords(masks, orginalText, out needRemoveSingleWord);
//合并到结果序列的对应位置中
if (unknownWords.Count > 0)
{
SuperLinkedListNode <WordInfo> cur = result.First;
if (needRemoveSingleWord && !_Options.ForceSingleWord)
{
//Remove single word need be remvoed
while (cur != null)
{
if (cur.Value.Word.Length == 1)
{
if (masks[cur.Value.Position] == 11)
{
SuperLinkedListNode <WordInfo> removeItem = cur;
cur = cur.Next;
result.Remove(removeItem);
continue;
}
}
cur = cur.Next;
}
}
cur = result.First;
j = 0;
while (cur != null)
{
if (cur.Value.Position >= unknownWords[j].Position)
{
result.AddBefore(cur, unknownWords[j]);
j++;
if (j >= unknownWords.Count)
{
break;
}
}
if (cur.Value.Position < unknownWords[j].Position)
{
cur = cur.Next;
}
}
while (j < unknownWords.Count)
{
result.AddLast(unknownWords[j]);
j++;
}
}
#endregion
return(result);
}