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);
}