public bool Match(string[] words, int startIdx, out int len, Gazetteer gazetteer)
{
int idx = startIdx;
len = 0;
if (string.Compare(words[idx], mWords[0], gazetteer.mIgnoreCase) != 0)
{
return(false);
} // first word must match
Utils.CaseType caseType = Utils.GetCaseType(words[idx]);
if (!(caseType == Utils.CaseType.ABC || caseType == Utils.CaseType.Abc || caseType == Utils.CaseType.AbC))
{
return(false);
} // *** only for the demo
idx++;
for (int i = 1; i < mWords.Count; i++)
{
while (idx < words.Length && gazetteer.IsStopWord(words[idx].ToLower()))
{
idx++;
} // skip stop words
if (idx == words.Length)
{
return(false);
}
if (string.Compare(words[idx++], mWords[i], gazetteer.mIgnoreCase) != 0)
{
return(false);
}
}
len = idx - startIdx;
return(true);
}
private static bool GetAcronymLemma(string word, string lemma, out string acronymLemma)
{
Match m = mAcronymRegex.Match(word);
if (m.Success && mLemListSuffix.Contains(m.Result("${suffix}").TrimStart('-', '–', '—')))
{
string acronym = m.Result("${acronym}");
Utils.CaseType caseType = Utils.GetCaseType(acronym);
if ((caseType == Utils.CaseType.Abc || caseType == Utils.CaseType.AbC) && lemma.Length >= 1)
{
acronymLemma = char.ToUpper(lemma[0]) + lemma.Substring(1);
return(true);
}
else if (caseType == Utils.CaseType.ABC) // uppercase
{
m = mAcronymRegex.Match(lemma);
if (m.Success)
{
acronymLemma = acronym + m.Result("${suffix}");
}
else
{
acronymLemma = acronym;
}
return(true);
}
}
acronymLemma = null;
return(false);
}
private static string ApplyLemmaRules(string lemma, string word, string tag)
{
if (tag == "N" || tag == "M")
{
return(word.ToLower());
}
if (tag == "O")
{
if (word.Length == 2 && word[1] == '.')
{
return(word);
}
else
{
return(word.ToLower());
}
}
if (word.Length >= 1)
{
Utils.CaseType caseType = Utils.GetCaseType(word);
bool isFirstCap = caseType == Utils.CaseType.Abc || caseType == Utils.CaseType.AbC || caseType == Utils.CaseType.ABC;
bool isAllCaps = caseType == Utils.CaseType.ABC;
if (tag.StartsWith("R"))
{
string acronymLemma;
if (GetAcronymLemma(word, lemma, out acronymLemma))
{
return(acronymLemma);
}
}
else if (tag.StartsWith("Kr"))
{
if (isAllCaps)
{
return(lemma.ToUpper());
}
}
else if (tag.StartsWith("Pp"))
{
string acronymLemma;
if (GetAcronymLemma(word, lemma, out acronymLemma))
{
return(acronymLemma);
}
else if (mLemListPpLemma.Contains(lemma))
{
return(char.ToUpper(lemma[0]) + lemma.Substring(1));
}
}
else if (tag.StartsWith("Ps"))
{
string acronymLemma;
if (GetAcronymLemma(word, lemma, out acronymLemma))
{
return(acronymLemma);
}
else if (mLemListPsLemma.Contains(lemma))
{
return(lemma);
}
else if (isFirstCap && lemma.Length >= 1)
{
return(char.ToUpper(lemma[0]) + lemma.Substring(1));
}
}
else if (tag.StartsWith("So"))
{
if (word.Length == 1 || mLemListSoLemma.Contains(word))
{
return(word);
}
Match m = mAcronymRegex.Match(word);
if (m.Success && mLemListSoLemma.Contains(m.Result("${acronym}")))
{
Match mLemma = mAcronymRegex.Match(lemma);
if (mLemma.Success)
{
return(m.Result("${acronym}") + mLemma.Result("${suffix}"));
}
else
{
return(m.Result("${acronym}"));
}
}
}
else if (tag.StartsWith("Sl"))
{
string acronymLemma;
if (GetAcronymLemma(word, lemma, out acronymLemma))
{
return(acronymLemma);
}
else if (isAllCaps)
{
return(lemma.ToUpper());
}
else if (isFirstCap && lemma.Length >= 1)
{
return(char.ToUpper(lemma[0]) + lemma.Substring(1));
}
}
}
return(lemma);
}
