// 全文中提取公司
public string extractCompany_fuzzy(string line)
{
line = line.ToLower();
string company = "";
var pattern = new Regex("(公\\s*司)(:|:)(([\u4e00-\u9fa5]|[a-z0-9])+)($|#|\\s)",
RegexOptions.IgnoreCase);
var matcher = pattern.Match(line);
if (matcher.Success)
{
company = matcher.Groups[3].Value;
return(company);
}
// TODO:交付时顺序后移到正则之后做召回 find maximum length position
List <string> companies = CompanyTokenAnalyzer.segment(line);
if (companies.Count > 0)
{
return(companies[0].ToUpper());
}
pattern = new Regex(
"(?!于|在|#)[\u4e00-\u9fa5]+?[\u4e00-\u9fa5|(|)|0-9|（|）]{2,20}(公司|集团|机构|代表处|办事处|营业部|经营部|事务所|学校|中心|研究所|研究院|酒店|商行|工作室|银行|俱乐部|加盟店|集团|门诊部)",
RegexOptions.IgnoreCase);
matcher = pattern.Match(line);
if (matcher.Success)
{
company = matcher.Groups[0].Value;
return(company);
}
return(company);
}
// 提取公司
public string extractCompany_exact(string line)
{
string company = "";
var pattern = new Regex("(公\\s*司)(:|:)(([\u4e00-\u9fa5]|[a-z0-9])+)($|#|\\s)",
RegexOptions.IgnoreCase);
var matcher = pattern.Match(line);
if (matcher.Success)
{
company = matcher.Groups[3].Value;
return(company);
}
pattern = new Regex(
"(?!于|在|#)[\u4e00-\u9fa5]+?[\u4e00-\u9fa5|(|)|0-9|（|）]{2,15}(公司|集团|代表处|办事处|营业部|经营部|事务所|学校|中心|研究所|研究院|酒店|商行|工作室|银行|俱乐部|加盟店|集团|门诊部)+",
RegexOptions.IgnoreCase);
matcher = pattern.Match(line);
if (matcher.Success)
{
company = matcher.Groups[0].Value;
return(company);
}
List <string> companies = CompanyTokenAnalyzer.segment(line);
if (companies.Count > 0)
{
return(companies[0]);
}
return(company);
}
