/// <summary>
/// 获得乙方
/// </summary>
/// <returns></returns>
string GetYiFang()
{
var Extractor = new ExtractPropertyByText();
//这些关键字后面
Extractor.LeadingColonKeyWordList = new string[] { "乙方：" };
//"供应商名称：","中标单位：","中标人：","中标单位：","中标人：","乙方（供方）：","承包人：","承包方：","中标方：","供应商名称：","中标人名称："
Extractor.ExtractFromTextFile(TextFileName);
foreach (var item in Extractor.CandidateWord)
{
var YiFang = item.Value.Trim();
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("乙方候补词(关键字)：[" + YiFang + "]");
}
return(YiFang);
}
//乙方:"有限公司"
//如果有子公司的话，优先使用子公司
foreach (var c in companynamelist)
{
if (c.isSubCompany)
{
return(c.secFullName);
}
}
var ExtractorHTML = new ExtractPropertyByHTML();
//这些关键字后面
ExtractorHTML.TrailingWordList = new string[] { "有限公司董事会" };
ExtractorHTML.Extract(root);
ExtractorHTML.CandidateWord.Reverse();
foreach (var item in ExtractorHTML.CandidateWord)
{
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("乙方候补词(关键字)：[" + item.Value.Trim() + "有限公司]");
}
return(item.Value.Trim() + "有限公司");
}
return(AnnouceCompanyName);
}
/// <summary>
/// 获得甲方
/// </summary>
/// <returns></returns>
public string GetJiaFang()
{
//最高置信度的抽取
EntityProperty e = new EntityProperty();
e.ExcludeContainsWordList = new string[] { "招标代理" };
e.LeadingColonKeyWordList = new string[] {
"甲方：", "合同买方：",
"发包人：", "发包单位：", "发包方：", "发包机构：", "发包人名称：",
"招标人：", "招标单位：", "招标方：", "招标机构：", "招标人名称：",
"业主：", "业主单位：", "业主方：", "业主机构：", "业主名称：",
"采购单位：", "采购单位名称：", "采购人：", "采购人名称：", "采购方：", "采购方名称："
};
e.CandidatePreprocess = (x =>
{
x = Normalizer.ClearTrailing(x);
return(CompanyNameLogic.AfterProcessFullName(x).secFullName);
});
e.MaxLength = ContractTraning.JiaFangES.MaxLength;
e.MaxLengthCheckPreprocess = Utility.TrimEnglish;
e.MinLength = 3;
e.Extract(this);
//这里不直接做Distinct，出现频次越高，则可信度越高
//多个甲方的时候，可能意味着没有甲方！
if (e.LeadingColonKeyWordCandidate.Distinct().Count() > 1)
{
foreach (var candidate in e.LeadingColonKeyWordCandidate)
{
Program.Logger.WriteLine("发现多个甲方：" + candidate);
}
}
if (e.LeadingColonKeyWordCandidate.Count > 0)
{
return(e.LeadingColonKeyWordCandidate[0]);
}
//招标
var Extractor = new ExtractPropertyByHTML();
var CandidateWord = new List <String>();
var StartArray = new string[] { "招标单位", "业主", "收到", "接到" };
var EndArray = new string[] { "发来", "发出", "的中标" };
Extractor.StartEndFeature = Utility.GetStartEndStringArray(StartArray, EndArray);
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var JiaFang = CompanyNameLogic.AfterProcessFullName(item.Value.Trim());
if (JiaFang.secFullName.Contains("招标代理"))
{
continue; //特殊业务规则
}
JiaFang.secFullName = JiaFang.secFullName.Replace("业主", String.Empty).Trim();
JiaFang.secFullName = JiaFang.secFullName.Replace("招标单位", String.Empty).Trim();
if (Utility.TrimEnglish(JiaFang.secFullName).Length > ContractTraning.JiaFangES.MaxLength)
{
continue;
}
if (JiaFang.secFullName.Length < 3)
{
continue; //使用实际长度排除全英文的情况
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("甲方候补词(招标)：[" + JiaFang.secFullName + "]");
}
CandidateWord.Add(JiaFang.secFullName);
}
//合同
Extractor = new ExtractPropertyByHTML();
StartArray = new string[] { "与", "与业主" };
EndArray = new string[] { "签署", "签订" };
Extractor.StartEndFeature = Utility.GetStartEndStringArray(StartArray, EndArray);
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var JiaFang = CompanyNameLogic.AfterProcessFullName(item.Value.Trim());
JiaFang.secFullName = JiaFang.secFullName.Replace("业主", String.Empty).Trim();
if (JiaFang.secFullName.Contains("招标代理"))
{
continue; //特殊业务规则
}
if (Utility.TrimEnglish(JiaFang.secFullName).Length > ContractTraning.JiaFangES.MaxLength)
{
continue;
}
if (JiaFang.secFullName.Length < 3)
{
continue; //使用实际长度排除全英文的情况
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("甲方候补词(合同)：[" + JiaFang.secFullName + "]");
}
CandidateWord.Add(JiaFang.secFullName);
}
return(CompanyNameLogic.MostLikeCompanyName(CandidateWord));
}
/// <summary>
/// 获得工程名
/// </summary>
/// <returns></returns>
string GetProjectName()
{
var ExtractorText = new ExtractPropertyByText();
//这些关键字后面(最优先)
ExtractorText.LeadingColonKeyWordList = new string[] { "项目名称：", "工程名称：", "中标项目：", "合同标的：", "工程内容：" };
ExtractorText.ExtractFromTextFile(TextFileName);
foreach (var item in ExtractorText.CandidateWord)
{
var ProjectName = item.Value.Trim();
if (EntityWordAnlayzeTool.TrimEnglish(ProjectName).Length > ContractTraning.MaxContractNameLength)
{
continue;
}
if (TrimJianCheng(ProjectName) == String.Empty)
{
continue;
}
ProjectName = TrimJianCheng(ProjectName);
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("项目名称候补词(关键字)：[" + ProjectName + "]");
}
return(ProjectName);
}
var Extractor = new ExtractPropertyByHTML();
Extractor.LeadingColonKeyWordList = ExtractorText.LeadingColonKeyWordList;
foreach (var item in Extractor.CandidateWord)
{
var ProjectName = item.Value.Trim();
if (EntityWordAnlayzeTool.TrimEnglish(ProjectName).Length > ContractTraning.MaxContractNameLength)
{
continue;
}
if (TrimJianCheng(ProjectName) == String.Empty)
{
continue;
}
ProjectName = TrimJianCheng(ProjectName);
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("项目名称候补词(关键字)：[" + ProjectName + "]");
}
return(ProjectName);
}
foreach (var bracket in quotationList)
{
if (bracket.Value.EndsWith("工程") ||
bracket.Value.EndsWith("标段"))
{
return(bracket.Value);
}
}
var MarkFeature = new ExtractPropertyByHTML.struMarkFeature();
MarkFeature.MarkStartWith = "“";
MarkFeature.MarkEndWith = "”";
MarkFeature.InnerEndWith = "标段";
var MarkFeatureConfirm = new ExtractPropertyByHTML.struMarkFeature();
MarkFeatureConfirm.MarkStartWith = "“";
MarkFeatureConfirm.MarkEndWith = "”";
MarkFeatureConfirm.InnerEndWith = "标";
Extractor.MarkFeature = new ExtractPropertyByHTML.struMarkFeature[] { MarkFeature, MarkFeatureConfirm };
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var ProjectName = item.Value.Trim();
if (EntityWordAnlayzeTool.TrimEnglish(ProjectName).Length > ContractTraning.MaxContractNameLength)
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("工程名称候补词（《XXX》）：[" + item + "]");
}
return(ProjectName);
}
var ExtractDP = new ExtractPropertyByDP();
var KeyList = new List <ExtractPropertyByDP.DPKeyWord>();
KeyList.Add(new ExtractPropertyByDP.DPKeyWord()
{
StartWord = new string[] { "确定为", "确定", "中标", "参与", "发布", "为" },
StartDPValue = new string[] { LTPTrainingDP.核心关系, LTPTrainingDP.定中关系, LTPTrainingDP.并列关系 },
EndWord = new string[] { "采购", "项目", "工程", "标段" },
EndDPValue = new string[] { }
});
ExtractDP.StartWithKey(KeyList, Dplist);
foreach (var item in ExtractDP.CandidateWord)
{
var ProjectName = item.Value.Trim();
if (EntityWordAnlayzeTool.TrimEnglish(ProjectName).Length > ContractTraning.MaxProjectNameLength)
{
continue;
}
if (ProjectName.Length <= 4)
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("工程候补词：[" + ProjectName + "]");
}
return(ProjectName);
}
return(String.Empty);
}
public void Extract(AnnouceDocument doc)
{
//纯关键字类型
if (KeyWordMap.Count != 0)
{
var candidate = ExtractByKeyWordMap(doc.root);
if (candidate.Count == 1)
{
WordMapResult = candidate.First();
}
if (candidate.Count > 1)
{
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("找到纯关键字类型两个关键字");
}
}
return;
}
if (LeadingColonKeyWordList != null)
{
//按照规则，由固定先导词的，例如 [项目名：]
//这里的词语不受任何其他因素制约，例如最大最小长度，有专用的预处理器
var ExtractorText = new ExtractPropertyByText();
//这些关键字后面:注意：TEXT版本可能存在空格，所以HTML版本也检查一遍
ExtractorText.LeadingColonKeyWordList = LeadingColonKeyWordList;
ExtractorText.ExtractFromTextFile(doc.TextFileName);
foreach (var item in ExtractorText.CandidateWord)
{
var PropertyValue = item.Value;
if (LeadingColonKeyWordCandidatePreprocess != null)
{
PropertyValue = LeadingColonKeyWordCandidatePreprocess(PropertyValue);
}
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
LeadingColonKeyWordCandidate.Add(PropertyValue);
}
var Extractor = new ExtractPropertyByHTML();
Extractor.LeadingColonKeyWordList = ExtractorText.LeadingColonKeyWordList;
Extractor.Extract(doc.root);
foreach (var item in ExtractorText.CandidateWord)
{
var PropertyValue = item.Value;
if (LeadingColonKeyWordCandidatePreprocess != null)
{
PropertyValue = LeadingColonKeyWordCandidatePreprocess(PropertyValue);
}
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
//TEXT里面有的，这里不重复添加了
if (!LeadingColonKeyWordCandidate.Contains(PropertyValue))
{
LeadingColonKeyWordCandidate.Add(PropertyValue);
}
}
}
//书名号和引号
if (QuotationTrailingWordList != null)
{
//接下来《》，“” 优先
foreach (var bracket in doc.quotationList)
{
foreach (var word in QuotationTrailingWordList)
{
if (bracket.Value.EndsWith(word))
{
var PropertyValue = CheckCandidate(bracket.Value);
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
QuotationTrailingCandidate.Add(PropertyValue);
}
}
}
}
//句法依存
if (DpKeyWordList != null)
{
var ExtractDP = new ExtractPropertyByDP();
ExtractDP.StartWithKey(DpKeyWordList, doc.Dplist);
foreach (var item in ExtractDP.CandidateWord)
{
var PropertyValue = CheckCandidate(item.Value);
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
DpKeyWordCandidate.Add(PropertyValue);
}
}
if (ExternalStartEndStringFeature != null)
{
var ExtractorTEXT = new ExtractPropertyByText();
ExtractorTEXT.StartEndFeature = ExternalStartEndStringFeature;
ExtractorTEXT.ExtractFromTextFile(doc.TextFileName);
foreach (var item in ExtractorTEXT.CandidateWord)
{
var PropertyValue = item.Value;
if (ExternalStartEndStringFeatureCandidatePreprocess != null)
{
PropertyValue = ExternalStartEndStringFeatureCandidatePreprocess(PropertyValue);
}
PropertyValue = CheckCandidate(PropertyValue);
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
ExternalStartEndStringFeatureCandidate.Add(PropertyValue);
}
//一部分无法提取TEXT的情况
var ExtractorHTML = new ExtractPropertyByHTML();
ExtractorHTML.StartEndFeature = ExternalStartEndStringFeature;
ExtractorHTML.Extract(doc.root);
foreach (var item in ExtractorHTML.CandidateWord)
{
var PropertyValue = item.Value;
if (ExternalStartEndStringFeatureCandidatePreprocess != null)
{
PropertyValue = ExternalStartEndStringFeatureCandidatePreprocess(PropertyValue);
}
PropertyValue = CheckCandidate(PropertyValue);
if (String.IsNullOrEmpty(PropertyValue))
{
continue;
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine(this.PropertyName + "：[" + PropertyValue + "]");
}
if (!ExternalStartEndStringFeatureCandidate.Contains(PropertyValue))
{
ExternalStartEndStringFeatureCandidate.Add(PropertyValue);
}
}
}
}
/// <summary>
/// 获得甲方
/// </summary>
/// <returns></returns>
string GetJiaFang(String YiFang)
{
//最高置信度的抽取
EntityProperty e = new EntityProperty();
e.ExcludeContainsWordList = new string[] { "招标代理" };
e.LeadingColonKeyWordList = new string[] {
"甲方：", "合同买方：",
"发包人：", "发包单位：", "发包方：", "发包机构：", "发包人名称：",
"招标人：", "招标单位：", "招标方：", "招标机构：", "招标人名称：", "项目招标人：",
"业主：", "业主单位：", "业主方：", "业主机构：", "业主名称：",
"采购单位：", "采购单位名称：", "采购人：", "采购人名称：", "采购方：", "采购方名称："
};
e.CandidatePreprocess = (x =>
{
x = Normalizer.ClearTrailing(x);
return(CompanyNameLogic.AfterProcessFullName(x).secFullName);
});
e.MaxLength = 32;
e.MaxLengthCheckPreprocess = Utility.TrimEnglish;
e.MinLength = 3;
e.Extract(this);
//这里不直接做Distinct，出现频次越高，则可信度越高
//多个甲方的时候，可能意味着没有甲方！
if (e.LeadingColonKeyWordCandidate.Distinct().Count() > 1)
{
foreach (var candidate in e.LeadingColonKeyWordCandidate)
{
Program.Logger.WriteLine("发现多个甲方：" + candidate);
}
}
if (e.LeadingColonKeyWordCandidate.Count > 0)
{
return(e.LeadingColonKeyWordCandidate[0]);
}
var ner = SearchJiaFang();
var NerJia = String.Empty;
if (!String.IsNullOrEmpty(ner))
{
foreach (var cn in companynamelist)
{
if (cn.secShortName == ner)
{
ner = cn.secFullName;
}
}
if (String.IsNullOrEmpty(YiFang))
{
NerJia = ner;
}
if (!YiFang.Equals(ner))
{
NerJia = ner;
}
}
//招标
var Extractor = new ExtractPropertyByHTML();
var CandidateWord = new List <String>();
var StartArray = new string[] { "招标单位", "业主", "收到", "接到" };
var EndArray = new string[] { "发来", "发出", "的中标" };
Extractor.StartEndFeature = Utility.GetStartEndStringArray(StartArray, EndArray);
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var JiaFang = CompanyNameLogic.AfterProcessFullName(item.Value.Trim());
if (JiaFang.secFullName.Contains("招标代理"))
{
continue; //特殊业务规则
}
JiaFang.secFullName = JiaFang.secFullName.Replace("业主", String.Empty).Trim();
JiaFang.secFullName = JiaFang.secFullName.Replace("招标单位", String.Empty).Trim();
if (Utility.TrimEnglish(JiaFang.secFullName).Length > 32)
{
continue;
}
if (JiaFang.secFullName.Length < 3)
{
continue; //使用实际长度排除全英文的情况
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("甲方候补词(招标)：[" + JiaFang.secFullName + "]");
}
CandidateWord.Add(JiaFang.secFullName);
}
//合同
Extractor = new ExtractPropertyByHTML();
StartArray = new string[] { "与", "与业主" };
EndArray = new string[] { "签署", "签订" };
Extractor.StartEndFeature = Utility.GetStartEndStringArray(StartArray, EndArray);
Extractor.Extract(root);
foreach (var item in Extractor.CandidateWord)
{
var JiaFang = CompanyNameLogic.AfterProcessFullName(item.Value.Trim());
JiaFang.secFullName = JiaFang.secFullName.Replace("业主", String.Empty).Trim();
if (JiaFang.secFullName.Contains("招标代理"))
{
continue; //特殊业务规则
}
if (Utility.TrimEnglish(JiaFang.secFullName).Length > 32)
{
continue;
}
if (JiaFang.secFullName.Length < 3)
{
continue; //使用实际长度排除全英文的情况
}
if (!Program.IsMultiThreadMode)
{
Program.Logger.WriteLine("甲方候补词(合同)：[" + JiaFang.secFullName + "]");
}
CandidateWord.Add(JiaFang.secFullName);
}
if (!String.IsNullOrEmpty(NerJia))
{
//原则上，有NER中提取的甲方，则使用甲方
foreach (var c in CandidateWord)
{
//但是，这里有可能是正确的解答，例如
//NER：(集团)有限公司 实际上应该是 XXXX(集团)有限公司
if (c.EndsWith(NerJia))
{
return(c);
}
}
return(NerJia);
}
else
{
return(CompanyNameLogic.MostLikeCompanyName(CandidateWord));
}
}