/// <summary>Tokenizes the string</summary>
/// <param name="input">The string to be tokenized</param>
/// <returns>A span array containing individual tokens as elements</returns>
public override Span[] TokenizePositions(string input)
{
if (string.IsNullOrEmpty(input))
{
return(new Span[0]);
}
var tokens = SplitOnWhitespaces(input);
var newTokens = new List <Span>();
var tokenProbabilities = new List <double>();
for (int currentToken = 0, tokenCount = tokens.Length; currentToken < tokenCount; currentToken++)
{
var tokenSpan = tokens[currentToken];
string token = input.Substring(tokenSpan.Start, (tokenSpan.End) - (tokenSpan.Start));
// Can't tokenize single characters
if (token.Length < 2)
{
newTokens.Add(tokenSpan);
tokenProbabilities.Add(1.0);
}
else if (AlphaNumericOptimization && AlphaNumeric.IsMatch(token))
{
newTokens.Add(tokenSpan);
tokenProbabilities.Add(1.0);
}
else
{
int startPosition = tokenSpan.Start;
int endPosition = tokenSpan.End;
int originalStart = tokenSpan.Start;
double tokenProbability = 1.0;
for (int currentPosition = originalStart + 1; currentPosition < endPosition; currentPosition++)
{
var context = _contextGenerator.GetContext(new Tuple <string, int>(token, currentPosition - originalStart));
double[] probabilities = _model.Evaluate(context);
string bestOutcome = _model.GetBestOutcome(probabilities);
tokenProbability *= probabilities[_model.GetOutcomeIndex(bestOutcome)];
if (bestOutcome == TokenContextGenerator.SplitIndicator)
{
newTokens.Add(new Span(startPosition, currentPosition));
tokenProbabilities.Add(tokenProbability);
startPosition = currentPosition;
tokenProbability = 1.0;
}
}
newTokens.Add(new Span(startPosition, endPosition));
tokenProbabilities.Add(tokenProbability);
}
}
return(newTokens.ToArray());
}
/// <summary>
/// Detect the position of the first words of sentences in a string.
/// </summary>
/// <param name="input">
/// The string to be processed.
/// </param>
/// <returns>
/// A integer array containing the positions of the end index of
/// every sentence
/// </returns>
public virtual int[] SentencePositionDetect(string input)
{
double sentenceProbability = 1;
_sentenceProbs.Clear();
var buffer = new StringBuilder(input);
List <int> endersList = _scanner.GetPositions(input);
var positions = new List <int>(endersList.Count);
for (int currentEnder = 0, enderCount = endersList.Count, index = 0; currentEnder < enderCount; currentEnder++)
{
int candidate = endersList[currentEnder];
int cInt = candidate;
// skip over the leading parts of non-token final delimiters
int firstWhiteSpace = GetFirstWhitespace(input, cInt + 1);
if (((currentEnder + 1) < enderCount) && ((endersList[currentEnder + 1]) < firstWhiteSpace))
{
continue;
}
var pair = new Tuple <StringBuilder, int>(buffer, candidate);
var context = _contextGenerator.GetContext(pair);
double[] probabilities = _model.Evaluate(context);
string bestOutcome = _model.GetBestOutcome(probabilities);
sentenceProbability *= probabilities[_model.GetOutcomeIndex(bestOutcome)];
if (bestOutcome.Equals("T") && IsAcceptableBreak(input, index, cInt))
{
if (index != cInt)
{
positions.Add(GetFirstNonWhitespace(input, GetFirstWhitespace(input, cInt + 1))); //moIntegerPool.GetInteger(GetFirstNonWhitespace(input, GetFirstWhitespace(input, cInt + 1))));
_sentenceProbs.Add(probabilities[_model.GetOutcomeIndex(bestOutcome)]);
}
index = cInt + 1;
}
}
return(positions.ToArray());
}
