public static void Anlayze(HTMLEngine.MyRootHtmlNode root, string KeyWord)
{
Console.WriteLine("关键字：[" + KeyWord + "]");
foreach (var paragrah in root.Children)
{
var segments = segmenter.Cut(paragrah.FullText).ToList(); // 默认为精确模式
Console.WriteLine("【精确模式】：{0}", string.Join("/ ", segments));
//寻找关键字的位置
for (int i = 0; i < segments.Count; i++)
{
if (segments[i].Equals(KeyWord))
{
//前5个词语和后五个词语
var startInx = Math.Max(0, i - 5);
var EndInx = Math.Min(i + 5, segments.Count);
for (int s = startInx; s < i; s++)
{
Console.WriteLine("前导关键字：[" + segments[s] + "]");
}
Console.WriteLine("关键字：[" + KeyWord + "]");
for (int s = i + 1; s < EndInx; s++)
{
Console.WriteLine("后续关键字：[" + segments[s] + "]");
}
}
}
}
}
public static List <LocAndValue <String> > LocatePercent(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue <String> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
var BracketList = RegularTool.GetChineseBrackets(OrgString);
Regex r = new Regex(RegularTool.PercentExpress);
foreach (var item in r.Matches(OrgString).ToList())
{
list.Add(new LocAndValue <String>()
{
Loc = sentence.PositionId,
Description = "百分比",
Value = item.Value,
StartIdx = item.Index
});
}
}
}
return(list);
}
//寻找同时含有关键字的列的表头
public static void PutTrainingItem(HTMLEngine.MyRootHtmlNode root, string KeyWord)
{
foreach (var Table in root.TableList)
{
var t = new HTMLTable(Table.Value);
for (int RowNo = 2; RowNo < t.RowCount; RowNo++)
{
//从第二行开始
for (int ColNo = 1; ColNo < t.ColumnCount; ColNo++)
{
if (t.CellValue(RowNo, ColNo).NormalizeKey().Equals(KeyWord.NormalizeKey()))
{
var title = t.CellValue(1, ColNo);
if (!TrainingTitleResult.ContainsKey(title))
{
TrainingTitleResult.Add(title, 1);
}
else
{
TrainingTitleResult[title]++;
}
}
}
}
}
}
/// <summary>
/// 增减持训练
/// </summary>
/// <param name="TraningCnt">训练条数</param>
public static void Traning(int TraningCnt = int.MaxValue)
{
var ChangeMethodTool = new TableAnlayzeTool();
var PreviewId = String.Empty;
var PreviewRoot = new HTMLEngine.MyRootHtmlNode();
int Cnt = 0;
foreach (var stockchange in TraningDataset.StockChangeList)
{
if (!PreviewId.Equals(stockchange.Id))
{
var htmlfile = Program.DocBase + @"\FDDC_announcements_round1_train_20180518\增减持\html\" + stockchange.Id + ".html";
PreviewRoot = new HTMLEngine().Anlayze(htmlfile, "");
PreviewId = stockchange.Id;
Cnt++; if (Cnt == TraningCnt)
{
break;
}
}
ChangeMethodTool.PutValueTrainingItem(PreviewRoot, new string[] { "减持方式", "增持方式" }.ToList());
}
var rank = Utility.FindTop(10, ChangeMethodTool.TrainingValueResult);
Program.Training.WriteLine("增减持方式");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
}
public static void TrainingIncreaseTarget()
{
TraningDataset.InitIncreaseStock();
var PreviewId = "";
var PreviewRoot = new HTMLEngine.MyRootHtmlNode();
foreach (var increase in TraningDataset.IncreaseStockList)
{
if (PreviewId.Equals(increase.id))
{
var htmlfile = Program.DocBase + @"\FDDC_announcements_round1_train_20180518\round1_train_20180518\定增\html\" + increase.id + ".html";
PreviewRoot = HTMLEngine.Anlayze(htmlfile);
}
TableAnlayzeTool.PutTrainingItem(PreviewRoot, increase.PublishTarget);
}
var Rank = new List <int>();
Rank = TableAnlayzeTool.TrainingTitleResult.Values.ToList();
Rank.Sort();
Rank.Reverse();
var Top10 = Rank[9];
foreach (var title in TableAnlayzeTool.TrainingTitleResult)
{
if (title.Value >= Top10)
{
Console.WriteLine(title.Key + ":" + title.Value);
}
}
}
/// <summary>
/// 自定义字符列表
/// </summary>
/// <param name="root"></param>
/// <param name="CustomerWord"></param>
/// <returns></returns>
public static List <LocAndValue <String> > LocateCustomerWord(HTMLEngine.MyRootHtmlNode root,
List <String> CustomerWord, string description = "字符")
{
var list = new List <LocAndValue <String> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
foreach (var word in CustomerWord)
{
int ScanStartIdx = 0;
while (OrgString.IndexOf(word, ScanStartIdx) != -1)
{
list.Add(new LocAndValue <String>()
{
Loc = sentence.PositionId,
Description = description,
Value = word,
StartIdx = OrgString.IndexOf(word, ScanStartIdx)
});
ScanStartIdx = OrgString.IndexOf(word, ScanStartIdx) + word.Length;
}
}
}
}
return(list);
}
/// <summary>
/// 寻找含有关键字的列的表头
/// </summary>
/// <param name="root"></param>
/// <param name="KeyWord"></param>
public void PutTitleTrainingItem(HTMLEngine.MyRootHtmlNode root, string KeyWord)
{
foreach (var Table in root.TableList)
{
var t = new HTMLTable(Table.Value);
for (int RowNo = 2; RowNo < t.RowCount; RowNo++)
{
//从第二行开始
for (int ColNo = 1; ColNo < t.ColumnCount; ColNo++)
{
var title = t.CellValue(1, ColNo).Replace(" ", "");
if (String.IsNullOrEmpty(title))
{
continue;
}
var value = t.CellValue(RowNo, ColNo);
if (Transform != null)
{
value = Transform(value, title);
}
if (value.NormalizeTextResult().Equals(KeyWord.NormalizeTextResult()))
{
if (!TrainingTitleResult.ContainsKey(title))
{
TrainingTitleResult.Add(title, 1);
}
else
{
TrainingTitleResult[title]++;
}
}
}
}
}
}
/// <summary>
/// 引号和书名号内容提取
/// </summary>
/// <param name="root">原始HTML</param>
/// <param name="IsSkipBracket">是否忽略括号内部的内容</param>
/// <returns></returns>
public static List <LocAndValue <String> > LocateQuotation(HTMLEngine.MyRootHtmlNode root, bool IsSkipBracket = true)
{
var list = new List <LocAndValue <String> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
var BracketList = RegularTool.GetChineseBrackets(OrgString);
Regex r = new Regex(@"\《.*?\》");
foreach (var item in r.Matches(OrgString).ToList())
{
bool IsContentInBracket = false;
foreach (var bracketItem in BracketList)
{
if (bracketItem.Contains(item.Value))
{
IsContentInBracket = true;
break;
}
}
if (IsSkipBracket && IsContentInBracket)
{
continue;
}
list.Add(new LocAndValue <String>()
{
Loc = sentence.PositionId,
Type = "字符",
Value = item.Value.Substring(1, item.Value.Length - 2)
});
}
r = new Regex(@"\“.*?\”");
foreach (var item in r.Matches(OrgString).ToList())
{
bool IsContentInBracket = false;
foreach (var bracketItem in BracketList)
{
if (bracketItem.Contains(item.Value))
{
IsContentInBracket = true;
break;
}
}
if (IsSkipBracket && IsContentInBracket)
{
continue;
}
list.Add(new LocAndValue <String>()
{
Loc = sentence.PositionId,
Type = "字符",
Value = item.Value.Substring(1, item.Value.Length - 2)
});
}
}
}
return(list);
}
//获得日期
public static List <LocAndValue <DateTime> > LocateDate(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue <DateTime> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
OrgString = DateUtility.ConvertUpperToLower(OrgString).Replace(" ", String.Empty);
var datelist = DateUtility.GetDate(OrgString);
foreach (var strDate in datelist)
{
var DateNumberList = RegularTool.GetNumberList(strDate);
String Year = DateNumberList[0];
String Month = DateNumberList[1];
String Day = DateNumberList[2];
int year; int month; int day;
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
list.Add(new LocAndValue <DateTime>()
{
Loc = sentence.PositionId,
Type = "日期",
Value = DateUtility.GetWorkDay(year, month, day)
});
}
}
}
}
return(list);
}
/// <summary>
/// 增发对象训练
/// </summary>
/// <param name="TraningCnt">训练条数</param>
public static void Training(int TraningCnt = int.MaxValue)
{
var TargetTool = new TableAnlayzeTool();
var IncreaseNumberTool = new TableAnlayzeTool();
IncreaseNumberTool.Transform = NumberUtility.NormalizerStockNumber;
var IncreaseMoneyTool = new TableAnlayzeTool();
IncreaseMoneyTool.Transform = MoneyUtility.Format;
var PreviewId = String.Empty;
var PreviewRoot = new HTMLEngine.MyRootHtmlNode();
int Cnt = 0;
foreach (var increase in TraningDataset.IncreaseStockList)
{
if (!PreviewId.Equals(increase.Id))
{
var htmlfile = Program.DocBase + @"\FDDC_announcements_round1_train_20180518\定增\html\" + increase.Id + ".html";
PreviewRoot = new HTMLEngine().Anlayze(htmlfile, "");
PreviewId = increase.Id;
Cnt++; if (Cnt == TraningCnt)
{
break;
}
}
TargetTool.PutTitleTrainingItem(PreviewRoot, increase.PublishTarget);
IncreaseNumberTool.PutTitleTrainingItem(PreviewRoot, increase.IncreaseNumber);
IncreaseMoneyTool.PutTitleTrainingItem(PreviewRoot, increase.IncreaseMoney);
}
var rank = Utility.FindTop(10, TargetTool.TrainingTitleResult);
Program.Training.WriteLine("增发对象");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
rank = Utility.FindTop(10, IncreaseNumberTool.TrainingTitleResult);
Program.Training.WriteLine("增发数量");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
rank = Utility.FindTop(10, IncreaseMoneyTool.TrainingTitleResult);
Program.Training.WriteLine("增发金额");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
}
public static List <struCompanyName> GetCompanyNameByCutWordFromHTML(HTMLEngine.MyRootHtmlNode root)
{
var namelist = new List <struCompanyName>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
GetCompany(namelist, sentence.Content, sentence.PositionId);
}
}
return(namelist);
}
public static string GetCompanyFullName(HTMLEngine.MyRootHtmlNode root)
{
var Extractor = new EntityProperty();
Extractor.TrailingWordList = new string[] { "公司董事会" };
Extractor.Extract(root);
Extractor.CandidateWord.Reverse();
foreach (var item in Extractor.CandidateWord)
{
Program.Logger.WriteLine("全称：[" + item + "公司]");
return(item);
}
return("");
}
//获得金额
public static List <LocAndValue> LocateMoney(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
OrgString = Utility.ConvertUpperDateToLittle(OrgString).Replace(" ", "");
var Money = Utility.SeekMoney(OrgString);
}
}
return(list);
}
List <string> ExtractByKeyWordMap(HTMLEngine.MyRootHtmlNode root)
{
var result = new List <string>();
foreach (var item in KeyWordMap)
{
var cnt = ExtractPropertyByHTML.FindWordCnt(item.Key, root).Count;
if (cnt > 0)
{
if (!result.Contains(item.Value))
{
result.Add(item.Value);
}
}
}
return(result);
}
List <string> ExtractByKeyWordMap(HTMLEngine.MyRootHtmlNode root)
{
var result = new List <string>();
foreach (var item in KeyWordMap)
{
var HasKey = ExtractPropertyByHTML.HasWord(item.Key, root);
if (HasKey)
{
if (!result.Contains(item.Value))
{
result.Add(item.Value);
}
}
}
return(result);
}
public static void GetEvaluateMethodTitle(int TraningCnt = int.MaxValue)
{
var TargetTool = new TableAnlayzeTool();
var PreviewId = String.Empty;
var PreviewRoot = new HTMLEngine.MyRootHtmlNode();
int Cnt = 0;
foreach (var ReOrg in TraningDataset.ReorganizationList)
{
if (!PreviewId.Equals(ReOrg.Id))
{
var htmlfile = Program.ReorganizationPath_TRAIN + Path.DirectorySeparatorChar + @"html" + Path.DirectorySeparatorChar + ReOrg.Id + ".html";
if (!System.IO.File.Exists(htmlfile))
{
continue;
}
PreviewRoot = new HTMLEngine().Anlayze(htmlfile, "");
PreviewId = ReOrg.Id;
Cnt++; if (Cnt == TraningCnt)
{
break;
}
}
if (!String.IsNullOrEmpty(ReOrg.EvaluateMethod))
{
TargetTool.PutTitleTrainingItemWithCodition(PreviewRoot, ReOrg.EvaluateMethod, ReOrg.TargetCompany);
}
}
var rank = Utility.FindTop(10, TargetTool.TrainingTitleResult);
Program.Training.WriteLine("评估方法");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
Program.Training.WriteLine("标的");
rank = Utility.FindTop(10, TargetTool.TrainingTitleCondition);
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
Program.Training.Flush();
}
/// <summary>
/// 寻找表中交易对手的标题
/// </summary>
/// <param name="TraningCnt"></param>
public static void GetTradeCompanyTitle(int TraningCnt = int.MaxValue)
{
var TargetTool = new TableAnlayzeTool();
var PreviewId = String.Empty;
var PreviewRoot = new HTMLEngine.MyRootHtmlNode();
int Cnt = 0;
foreach (var ReOrg in TraningDataset.ReorganizationList)
{
if (!PreviewId.Equals(ReOrg.Id))
{
var htmlfile = Program.ReorganizationPath_TRAIN + @"\html\" + ReOrg.Id + ".html";
if (!System.IO.File.Exists(htmlfile))
{
continue;
}
PreviewRoot = new HTMLEngine().Anlayze(htmlfile, "");
PreviewId = ReOrg.Id;
Cnt++; if (Cnt == TraningCnt)
{
break;
}
}
foreach (var item in ReOrg.TradeCompany.Split(Utility.SplitChar))
{
TargetTool.PutTitleTrainingItem(PreviewRoot, item);
}
}
var rank = Utility.FindTop(10, TargetTool.TrainingTitleResult);
Program.Training.WriteLine("交易对象");
foreach (var rec in rank)
{
Program.Training.WriteLine(rec.ToString());
}
foreach (var item in TargetTool.WholeHeaderRow)
{
Program.Training.WriteLine(item);
}
Program.Training.Flush();
}
/// <summary>
/// 股数
/// </summary>
/// <param name="root"></param>
/// <returns></returns>
public static List <LocAndValue <String> > LocateStockNumber(HTMLEngine.MyRootHtmlNode root)
{
var targetRegular = new ExtractProperyBase.struRegularExpressFeature()
{
RegularExpress = @"\d+(,\d+)+",
TrailingWordList = new string[] { "股" }.ToList()
};
var list = new List <LocAndValue <String> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var ExpResult = ExtractPropertyByHTML.RegularExFinder(sentence.PositionId, sentence.Content, targetRegular, "|");
list.AddRange(ExpResult);
}
}
return(list);
}
//固定搭配
public static string GetCompanyShortName(HTMLEngine.MyRootHtmlNode root)
{
var companyList = new Dictionary <string, string>();
//从第一行开始找到 有限公司 有限责任公司, 如果有简称的话Value是简称
//股票简称：东方电气
//东方电气股份有限公司董事会
var Extractor = new EntityProperty();
Extractor.LeadingWordList = new string[] { "股票简称", "证券简称" };
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var ShortName = item.Replace(":", "").Replace("：", "").Trim();
if (Utility.GetStringBefore(ShortName, "、") != "")
{
ShortName = Utility.GetStringBefore(ShortName, "、");
}
if (Utility.GetStringBefore(ShortName, "）") != "")
{
ShortName = Utility.GetStringBefore(ShortName, "）");
}
if (Utility.GetStringBefore(ShortName, "公告") != "")
{
ShortName = Utility.GetStringBefore(ShortName, "公告");
}
if (Utility.GetStringBefore(ShortName, "股票") != "")
{
ShortName = Utility.GetStringBefore(ShortName, "股票");
}
if (Utility.GetStringBefore(ShortName, "证券") != "")
{
ShortName = Utility.GetStringBefore(ShortName, "证券");
}
if (Utility.GetStringBefore(ShortName, " ") != "")
{
ShortName = Utility.GetStringBefore(ShortName, " ");
}
FDDC.Program.Logger.WriteLine("简称:[" + ShortName + "]");
return(ShortName);
}
return("");
}
/// <summary>
/// 自定义字符列表
/// </summary>
/// <param name="root"></param>
/// <param name="CustomerWord"></param>
/// <returns></returns>
public static List <LocAndValue <String> > LocateCustomerWord(HTMLEngine.MyRootHtmlNode root,
List <String> CustomerWord, string description = "字符")
{
var list = new List <LocAndValue <String> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content.Replace(" ", "");
foreach (var word in CustomerWord)
{
if (String.IsNullOrEmpty(word))
{
continue;
}
int ScanStartIdx = 0;
int Count = 0;
while (OrgString.IndexOf(word, ScanStartIdx) != -1)
{
list.Add(new LocAndValue <String>()
{
Loc = sentence.PositionId,
Description = description,
Value = word,
StartIdx = OrgString.IndexOf(word, ScanStartIdx)
});
Count++;
if (Count > 5000)
{
//死循环的防止
Console.WriteLine("OrgString:" + OrgString);
Console.WriteLine("word:[" + word + "]");
throw new System.ArgumentException();
}
ScanStartIdx = OrgString.IndexOf(word, ScanStartIdx) + word.Length;
}
}
}
}
return(list);
}
//获得日期
public static List <LocAndValue> LocateDate(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
OrgString = Utility.ConvertUpperDateToLittle(OrgString).Replace(" ", "");
if (!String.IsNullOrEmpty(RegularTool.GetDate(OrgString)))
{
list.Add(new LocAndValue()
{
Loc = sentence.PositionId, Value = RegularTool.GetDate(OrgString)
});
}
}
}
return(list);
}
public static void AnlayzeEntitySurroundWords(HTMLEngine.MyRootHtmlNode root, string KeyWord)
{
Program.Training.WriteLine("关键字：[" + KeyWord + "]");
JiebaSegmenter segmenter = new JiebaSegmenter();
segmenter.AddWord(KeyWord);
foreach (var paragrah in root.Children)
{
var segments = segmenter.Cut(paragrah.FullText.NormalizeKey()).ToList(); // 默认为精确模式
//Console.WriteLine("【精确模式】：{0}", string.Join("/ ", segments));
//寻找关键字的位置
for (int i = 0; i < segments.Count; i++)
{
if (segments[i].Equals(KeyWord))
{
//前5个词语和后五个词语
var startInx = Math.Max(0, i - 5);
var EndInx = Math.Min(i + 5, segments.Count);
for (int s = startInx; s < i; s++)
{
Program.Training.WriteLine("前导关键字：[" + segments[s] + "]");
if (segments[s] == "：")
{
var leading = "";
for (int l = startInx; l < s; l++)
{
leading += segments[l];
}
Console.WriteLine("冒号前导词：" + leading);
}
}
Program.Training.WriteLine("关键字：[" + KeyWord + "]");
for (int s = i + 1; s < EndInx; s++)
{
Program.Training.WriteLine("后续关键字：[" + segments[s] + "]");
}
}
}
}
}
/// <summary>
/// 某类标题的值
/// </summary>
/// <param name="root"></param>
/// <param name="KeyWord"></param>
public void PutValueTrainingItem(HTMLEngine.MyRootHtmlNode root, List <string> TitleKeyWord)
{
foreach (var Table in root.TableList)
{
var t = new HTMLTable(Table.Value);
for (int RowNo = 2; RowNo < t.RowCount; RowNo++)
{
//从第二行开始
for (int ColNo = 1; ColNo < t.ColumnCount; ColNo++)
{
var title = t.CellValue(1, ColNo).Replace(" ", "");
if (String.IsNullOrEmpty(title))
{
continue;
}
var value = t.CellValue(RowNo, ColNo).NormalizeTextResult();
if (string.IsNullOrEmpty(value))
{
continue;
}
foreach (var key in TitleKeyWord)
{
if (title.Equals(key))
{
if (!TrainingValueResult.ContainsKey(value))
{
TrainingValueResult.Add(value, 1);
}
else
{
TrainingValueResult[value]++;
}
}
}
}
}
}
}
//词法分析
public static List <String> GetProjectName(HTMLEngine.MyRootHtmlNode root)
{
var posSeg = new PosSegmenter();
var namelist = new List <String>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var words = posSeg.Cut(sentence.Content).ToList();
for (int baseInd = 0; baseInd < words.Count; baseInd++)
{
if (words[baseInd].Word == "标段" ||
words[baseInd].Word == "工程" ||
words[baseInd].Word == "项目")
{
var projectName = "";
//是否能够在前面找到地名
for (int NRIdx = baseInd; NRIdx > -1; NRIdx--)
{
//地理
if (words[NRIdx].Flag == "ns")
{
projectName = "";
for (int companyFullNameInd = NRIdx; companyFullNameInd <= baseInd; companyFullNameInd++)
{
projectName += words[companyFullNameInd].Word;
}
namelist.Add(projectName);
break; //不要继续寻找地名了
}
}
}
}
}
}
return(namelist);
}
//获得日期
public static List <LocAndValue <(DateTime StartDate, DateTime EndDate)> > LocateDateRange(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue <(DateTime StartDate, DateTime EndDate)> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
OrgString = DateUtility.ConvertUpperToLower(OrgString).Replace(" ", String.Empty);
var datelist = DateUtility.GetRangeDate(OrgString);
foreach (var strDate in datelist)
{
var DateNumberList = RegularTool.GetNumberList(strDate);
DateTime ST = new DateTime();
DateTime ED = new DateTime();
if (DateNumberList.Count == 6)
{
String Year = DateNumberList[0];
String Month = DateNumberList[1];
String Day = DateNumberList[2];
int year; int month; int day;
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ST = DateUtility.GetWorkDay(year, month, day);
}
Year = DateNumberList[3];
Month = DateNumberList[4];
Day = DateNumberList[5];
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ED = DateUtility.GetWorkDay(year, month, day);
}
list.Add(new LocAndValue <(DateTime StartDate, DateTime EndDate)>()
{
Loc = sentence.PositionId,
Type = "日期范围",
Value = (ST, ED)
});
}
if (DateNumberList.Count == 5)
{
String Year = DateNumberList[0];
String Month = DateNumberList[1];
String Day = DateNumberList[2];
int year; int month; int day;
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ST = DateUtility.GetWorkDay(year, month, day);
}
Month = DateNumberList[3];
Day = DateNumberList[4];
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ED = DateUtility.GetWorkDay(year, month, day);
}
list.Add(new LocAndValue <(DateTime StartDate, DateTime EndDate)>()
{
Loc = sentence.PositionId,
Type = "日期范围",
Value = (ST, ED)
});
}
if (DateNumberList.Count == 4)
{
String Year = DateNumberList[0];
String Month = DateNumberList[1];
String Day = DateNumberList[2];
int year; int month; int day;
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ST = DateUtility.GetWorkDay(year, month, day);
}
Day = DateNumberList[3];
if (int.TryParse(Year, out year) && int.TryParse(Month, out month) && int.TryParse(Day, out day))
{
ED = DateUtility.GetWorkDay(year, month, day);
}
list.Add(new LocAndValue <(DateTime StartDate, DateTime EndDate)>()
{
Loc = sentence.PositionId,
Type = "日期范围",
Value = (ST, ED)
});
}
}
}
}
return(list);
}
public static List <struCompanyName> GetCompanyNameByCutWord(HTMLEngine.MyRootHtmlNode root)
{
var posSeg = new PosSegmenter();
var namelist = new List <struCompanyName>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var words = posSeg.Cut(sentence.Content).ToList();
for (int baseInd = 0; baseInd < words.Count; baseInd++)
{
var FullName = "";
var ShortName = "";
if (words[baseInd].Word == "有限公司" ||
(words[baseInd].Word == "有限" && baseInd != words.Count - 1 && words[baseInd + 1].Word == "合伙"))
{
//是否能够在前面找到地名
for (int NRIdx = baseInd; NRIdx > -1; NRIdx--)
{
//地理
if (words[NRIdx].Flag == "ns")
{
FullName = "";
for (int companyFullNameInd = NRIdx; companyFullNameInd <= baseInd; companyFullNameInd++)
{
FullName += words[companyFullNameInd].Word;
}
if (words[baseInd].Word == "有限")
{
FullName += words[baseInd + 1].Word;
FullName += words[baseInd + 2].Word;
}
break; //不要继续寻找地名了
}
}
//是否能够在后面找到简称
for (int JCIdx = baseInd; JCIdx < words.Count; JCIdx++)
{
//地理
if (words[JCIdx].Word.Equals("简称"))
{
var ShortNameStart = -1;
var ShortNameEnd = -1;
for (int ShortNameIdx = baseInd; ShortNameIdx < words.Count; ShortNameIdx++)
{
if (words[ShortNameIdx].Word.Equals("“"))
{
ShortNameStart = ShortNameIdx + 1;
}
if (words[ShortNameIdx].Word.Equals("”"))
{
ShortNameEnd = ShortNameIdx - 1;
break;
}
}
if (ShortNameStart != -1 && ShortNameEnd != -1)
{
ShortName = "";
for (int i = ShortNameStart; i <= ShortNameEnd; i++)
{
ShortName += words[i].Word;
}
}
}
}
if (FullName != "")
{
namelist.Add(new struCompanyName()
{
secFullName = FullName, secShortName = ShortName
});
}
}
}
}
}
return(namelist);
}
//获得金额
public static List <LocAndValue <(String MoneyAmount, String MoneyCurrency)> > LocateMoney(HTMLEngine.MyRootHtmlNode root)
{
var list = new List <LocAndValue <(String MoneyAmount, String MoneyCurrency)> >();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
var OrgString = sentence.Content;
OrgString = MoneyUtility.ConvertUpperToLower(OrgString).Replace(" ", String.Empty);
var Money = MoneyUtility.SeekMoney(OrgString);
foreach (var money in Money)
{
list.Add(new LocAndValue <(String MoneyAmount, String MoneyCurrency)>
{
Loc = sentence.PositionId,
Type = "金额",
Value = money
});
}
}
}
return(list);
}
public static List <struCompanyName> GetCompanyNameByCutWord(HTMLEngine.MyRootHtmlNode root)
{
var posSeg = new PosSegmenter();
var namelist = new List <struCompanyName>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
if (string.IsNullOrEmpty(sentence.Content))
{
continue;
}
var words = posSeg.Cut(sentence.Content).ToList();
var PreviewEndIdx = -1;
for (int baseInd = 0; baseInd < words.Count; baseInd++)
{
var FullName = "";
var ShortName = "";
var IsSubCompany = false;
var StartIdx = -1;
if (
words[baseInd].Word == "有限公司" ||
(words[baseInd].Word == "公司" && baseInd != 0 && words[baseInd - 1].Word == "承包") ||
(words[baseInd].Word == "有限" && baseInd != words.Count - 1 && words[baseInd + 1].Word == "合伙")
)
{
//是否能够在前面找到地名
for (int NRIdx = baseInd; NRIdx > PreviewEndIdx; NRIdx--)
{
//地理
if (words[NRIdx].Flag == EntityWordAnlayzeTool.地名)
{
FullName = "";
for (int companyFullNameInd = NRIdx; companyFullNameInd <= baseInd; companyFullNameInd++)
{
FullName += words[companyFullNameInd].Word;
}
//承包公司
if (words[baseInd].Word == "公司")
{
//什么都不用做
}
//(有限合伙)
if (words[baseInd].Word == "有限")
{
FullName += words[baseInd + 1].Word;
FullName += words[baseInd + 2].Word;
}
//子公司判断
if (NRIdx != 0 && words[NRIdx - 1].Word == "子公司")
{
IsSubCompany = true;
}
StartIdx = NRIdx;
PreviewEndIdx = baseInd;
break; //不要继续寻找地名了
}
}
//是否能够在后面找到简称
for (int JCIdx = baseInd; JCIdx < words.Count; JCIdx++)
{
//地理
if (words[JCIdx].Word.Equals("简称"))
{
var ShortNameStart = -1;
var ShortNameEnd = -1;
for (int ShortNameIdx = baseInd; ShortNameIdx < words.Count; ShortNameIdx++)
{
if (words[ShortNameIdx].Word.Equals("“"))
{
ShortNameStart = ShortNameIdx + 1;
}
if (words[ShortNameIdx].Word.Equals("”"))
{
ShortNameEnd = ShortNameIdx - 1;
break;
}
}
if (ShortNameStart != -1 && ShortNameEnd != -1)
{
ShortName = "";
for (int i = ShortNameStart; i <= ShortNameEnd; i++)
{
ShortName += words[i].Word;
}
}
}
}
if (FullName != "")
{
namelist.Add(new struCompanyName()
{
secFullName = FullName,
secShortName = ShortName,
isSubCompany = IsSubCompany,
positionId = sentence.PositionId,
WordIdx = StartIdx
});
}
}
}
}
}
return(namelist);
}
/// <summary>
/// 带条件的标题检索
/// </summary>
/// <param name="root"></param>
/// <param name="KeyWord"></param>
/// <param name="ConditionKey"></param>
public void PutTitleTrainingItemWithCodition(HTMLEngine.MyRootHtmlNode root, string KeyWord, string ConditionKey)
{
if (root.TableList == null)
{
return;
}
foreach (var Table in root.TableList)
{
var t = new HTMLTable(Table.Value);
for (int RowNo = 2; RowNo < t.RowCount; RowNo++)
{
var IsConditionOK = false;
var ConditionTitle = "";
for (int ColNo = 1; ColNo < t.ColumnCount; ColNo++)
{
var title = t.CellValue(1, ColNo).Replace(" ", "");
if (String.IsNullOrEmpty(title))
{
continue;
}
var value = t.CellValue(RowNo, ColNo);
if (value.NormalizeTextResult().Contains(ConditionKey.NormalizeTextResult()))
{
ConditionTitle = title;
IsConditionOK = true;
break;
}
}
if (!IsConditionOK)
{
continue;
}
//从第二行开始
for (int ColNo = 1; ColNo < t.ColumnCount; ColNo++)
{
var title = t.CellValue(1, ColNo).Replace(" ", "");
if (String.IsNullOrEmpty(title))
{
continue;
}
var value = t.CellValue(RowNo, ColNo);
if (Transform != null)
{
value = Transform(value, title);
}
if (value.NormalizeTextResult().Equals(KeyWord.NormalizeTextResult()))
{
if (!TrainingTitleResult.ContainsKey(title))
{
TrainingTitleResult.Add(title, 1);
}
else
{
TrainingTitleResult[title]++;
}
if (!TrainingTitleCondition.ContainsKey(ConditionTitle))
{
TrainingTitleCondition.Add(ConditionTitle, 1);
}
else
{
TrainingTitleCondition[ConditionTitle]++;
}
}
}
}
}
}
public static List <struCompanyName> GetCompanyNameByCutWord(HTMLEngine.MyRootHtmlNode root)
{
var posSeg = new PosSegmenter();
var namelist = new List <struCompanyName>();
foreach (var paragrah in root.Children)
{
foreach (var sentence in paragrah.Children)
{
if (string.IsNullOrEmpty(sentence.Content))
{
continue;
}
var words = posSeg.Cut(sentence.Content).ToList();
var PreviewEndIdx = -1;
for (int baseInd = 0; baseInd < words.Count; baseInd++)
{
var FullName = String.Empty;
var ShortName = String.Empty;
var IsSubCompany = false;
if (words[baseInd].Word == "国家电网" &&
(baseInd + 1) < words.Count &&
words[baseInd + 1].Word == "公司")
{
namelist.Add(new struCompanyName()
{
secFullName = "国家电网公司",
positionId = sentence.PositionId,
WordIdx = baseInd,
Score = 100
});
continue;
}
if (
words[baseInd].Word == "有限公司" ||
(words[baseInd].Word == "公司" && baseInd != 0 && words[baseInd - 1].Word == "有限责任") ||
(words[baseInd].Word == "公司" && baseInd != 0 && words[baseInd - 1].Word == "承包") ||
(words[baseInd].Word == "有限" && baseInd != words.Count - 1 && words[baseInd + 1].Word == "合伙")
)
{
//是否能够在后面找到简称
for (int JCIdx = baseInd; JCIdx < words.Count; JCIdx++)
{
//简称关键字
if (words[JCIdx].Word.Equals("简称") || words[JCIdx].Word.Equals("称"))
{
var ShortNameStart = -1;
var ShortNameEnd = -1;
for (int ShortNameIdx = JCIdx; ShortNameIdx < words.Count; ShortNameIdx++)
{
if (words[ShortNameIdx].Word.Equals("“"))
{
ShortNameStart = ShortNameIdx + 1;
}
if (words[ShortNameIdx].Word.Equals("”"))
{
ShortNameEnd = ShortNameIdx - 1;
break;
}
}
if (ShortNameStart != -1 && ShortNameEnd != -1)
{
ShortName = String.Empty;
for (int i = ShortNameStart; i <= ShortNameEnd; i++)
{
ShortName += words[i].Word;
}
}
break;
}
}
var FirstShortNameWord = String.Empty;
if (ShortName.Length == 4)
{
FirstShortNameWord = ShortName.Substring(0, 2);
}
var IsMarkClosed = true;
var CompanyStartIdx = -1;
var FirstShortNameIdx = -1; //包含简称的位置
//是否能够在前面找到地名
for (int NRIdx = baseInd; NRIdx > PreviewEndIdx; NRIdx--)
{
if (words[NRIdx].Word == FirstShortNameWord)
{
FirstShortNameIdx = NRIdx; //备用
}
//寻找地名?words[NRIdx].Flag == EntityWordAnlayzeTool.机构团体
//posSeg.Cut(words[NRIdx].Word + "市").First().Flag == EntityWordAnlayzeTool.地名
if (words[NRIdx].Flag == LTPTrainingNER.地名 || PosNS.NsDict.Contains(words[NRIdx].Word))
{
//注意，地名可能相连，例如：上海市嘉定
if (NRIdx != 0 && (words[NRIdx - 1].Flag == LTPTrainingNER.地名 || PosNS.NsDict.Contains(words[NRIdx - 1].Word)))
{
continue;
}
FullName = String.Empty;
for (int companyFullNameInd = NRIdx; companyFullNameInd <= baseInd; companyFullNameInd++)
{
FullName += words[companyFullNameInd].Word;
}
//(有限合伙)
if (words[baseInd].Word == "有限")
{
FullName += words[baseInd + 1].Word;
if ((baseInd + 2) < words.Count)
{
FullName += words[baseInd + 2].Word;
}
}
//子公司判断
if (NRIdx != 0 && words[NRIdx - 1].Word == "子公司")
{
IsSubCompany = true;
}
if (IsMarkClosed)
{
//皆大欢喜的局面
CompanyStartIdx = NRIdx;
PreviewEndIdx = baseInd;
break; //不要继续寻找地名了
}
}
if (words[NRIdx].Flag == LTPTrainingNER.词性标点)
{
if (words[NRIdx].Word != "（" && words[NRIdx].Word != "）")
{
break;
}
if (words[NRIdx].Word == "）")
{
IsMarkClosed = false; //打开
}
if (words[NRIdx].Word == "（")
{
IsMarkClosed = true; //关闭
}
}
}
if (CompanyStartIdx == -1)
{
if (FirstShortNameIdx == -1)
{
continue;
}
if (posSeg.Cut(ShortName).First().Flag == LTPTrainingNER.地名)
{
continue;
}
FullName = String.Empty;
for (int NRIdx = FirstShortNameIdx; NRIdx <= baseInd; NRIdx++)
{
FullName += words[NRIdx].Word;
}
//(有限合伙)
if (words[baseInd].Word == "有限")
{
FullName += words[baseInd + 1].Word;
FullName += words[baseInd + 2].Word;
}
//子公司判断
if (FirstShortNameIdx != 0 && words[FirstShortNameIdx - 1].Word == "子公司")
{
IsSubCompany = true;
}
}
if (FullName != String.Empty)
{
FullName = FullName.Replace(" ", String.Empty).Trim();
ShortName = ShortName.Replace(" ", String.Empty).Trim();
if (ShortName == "公司" || ShortName == "本公司")
{
ShortName = String.Empty;
}
if (ShortName == String.Empty)
{
var json = GetCompanyNameByFullName(FullName);
ShortName = json.secShortName;
}
namelist.Add(new struCompanyName()
{
secFullName = FullName,
secShortName = ShortName,
isSubCompany = IsSubCompany,
positionId = sentence.PositionId,
WordIdx = CompanyStartIdx,
Score = 100
});
}
}
}
}
}
return(namelist);
}
