/// <summary>
/// This method splits the input word by adding a space after ending
/// punctuation. The input must be single word (no space).
/// The process method splits the inWord by adding space(s) after endingPunc.
/// Current algorithm can only handle max. up to 3 endignPuncs.
/// One in each component of coreTermObj: coreTerm, prefix, and suffix.
/// - prefix: leading str with punc|spac|number
/// - coreterm: = the original str - prefix - suffix
/// - suffix: ending str with punc|space|number
/// This can be improved by using recursive algorithm in the coreTerm.
/// For example: "ankle,before.The" in 15737.txt will be split twice in
/// recursive algorithm.
/// </summary>
/// <param name="inWord"> the input token (single word)
/// </param>
/// <returns> the splited word. </returns>
public static string Process(string inWord)
{
string outWord = inWord;
bool debugFlag = false;
// eProcess: check if can skip
int ctType = CoreTermUtil.CT_TYPE_SPACE_PUNC_DIGIT;
if (IsQualified(inWord) == true)
{
// 0. convert to coreTerm object
bool splitFlag = false;
CoreTermObj cto = new CoreTermObj(inWord, ctType);
// 1. update coreTerm
string inCoreTerm = cto.GetCoreTerm();
string lastEndingPunc = FindLastEndingPunc(inCoreTerm);
// add a space after the last endingPunc
if (!string.ReferenceEquals(lastEndingPunc, null))
{
// get the splitObj and then the split string
string outCoreTerm = EndingPunc.GetSplitStr(inCoreTerm, lastEndingPunc);
cto.SetCoreTerm(outCoreTerm);
splitFlag = true;
}
// 2. update the prefix when it ends with a endingPunc
// prefix contains punc and numbers
string prefix = cto.GetPrefix();
if ((prefix.Length != 0) && (EndsWithEndingPunc(prefix) == true)) // ends with endingPunc
{
prefix = prefix + GlobalVars.SPACE_STR;
cto.SetPrefix(prefix);
splitFlag = true;
}
// 3. update the suffix and add a space after the last endingPunc
// suffix contians punctuation and numbers
string suffix = cto.GetSuffix();
if ((suffix.Length != 0) && (ContainsEndingPunc(suffix) == true) && (IsPureEndingPunc(suffix) == false)) // can't be pure endingPuncs
// add space after the last endingPunc
{
string lastEndingPunc2 = FindLastEndingPunc(suffix);
if (!string.ReferenceEquals(lastEndingPunc2, null))
{
// get the splitObj and then the split string
string outSuffix = EndingPunc.GetSplitStr(suffix, lastEndingPunc2);
cto.SetSuffix(outSuffix);
splitFlag = true;
}
}
// update outWord
if (splitFlag == true)
{
outWord = cto.ToString();
}
}
return(outWord);
}
private static void TestExceptionCA()
{
Console.WriteLine("----- Ending Punc Exception: Comma -----");
List <string> inWordList = new List <string>();
inWordList.Add("50,000");
inWordList.Add("1,234,567");
inWordList.Add("123");
inWordList.Add("12,34");
foreach (string inWord in inWordList)
{
Console.WriteLine("- IsException(" + inWord + "): " + EndingPunc.IsException(inWord, ENDING_CA));
}
}
private static void TestExceptionEM()
{
Console.WriteLine("----- Ending Punc Exception: Exclamation Mark -----");
List <string> inWordList = new List <string>();
inWordList.Add("ulcers!'");
inWordList.Add("ulcers!\"");
inWordList.Add("ulcers!]");
inWordList.Add("XX!'test");
foreach (string inWord in inWordList)
{
Console.WriteLine("- IsException(" + inWord + "): " + EndingPunc.IsException(inWord, ENDING_EM));
}
}
private static void TestExceptionQM()
{
Console.WriteLine("----- Ending Punc Exception: Question Mark -----");
List <string> inWordList = new List <string>();
inWordList.Add("ulcers?'"); // 12769.txt
inWordList.Add("ulcers?\"");
inWordList.Add("ulcers?]");
inWordList.Add("XX?'test");
foreach (string inWord in inWordList)
{
Console.WriteLine("- IsException(" + inWord + "): " + EndingPunc.IsException(inWord, ENDING_QM));
}
}
private static void TestExceptionP()
{
Console.WriteLine("----- Ending Punc Exception: period -----");
List <string> inWordPList = new List <string>();
// exception 1
inWordPList.Add("Dr.s");
inWordPList.Add("Mr.s");
// exception 2
inWordPList.Add("16q22.1");
inWordPList.Add("123.2");
inWordPList.Add("123.234.4567");
inWordPList.Add("1c3.2d4.4e6");
inWordPList.Add("123.23a4.456");
inWordPList.Add("123a.234.456");
// exception 3
inWordPList.Add("D.C.A.B.");
inWordPList.Add("D.C.A.B");
inWordPList.Add("d.c.a.");
inWordPList.Add("d.c.a");
inWordPList.Add("D.c");
inWordPList.Add("D.CC.A.B.");
inWordPList.Add("DD.C.A.B.");
inWordPList.Add("d.1.a.");
inWordPList.Add("D.123.A.B.");
// exception 4
inWordPList.Add("St.-John");
inWordPList.Add("123.-John");
inWordPList.Add("#$.-John");
inWordPList.Add("St.$%^John");
inWordPList.Add("St.John");
inWordPList.Add("St.J.");
inWordPList.Add("Test...123");
foreach (string inWordP in inWordPList)
{
Console.WriteLine("- IsException(" + inWordP + "): " + EndingPunc.IsException(inWordP, ENDING_P));
}
}
