// For cases like "more than 30000 in 2010", we will not treate "30000 in 2010" as a fraction number
// In this method, "30000 in 2010" will be changed to "30000"
private List <ExtractResult> RemoveAmbiguousFractions(List <ExtractResult> ers)
{
foreach (var er in ers)
{
if (er.Data != null && er.Data.ToString() == RegexTagGenerator.GenerateRegexTag(Constants.FRACTION_PREFIX, Constants.ENGLISH).Name)
{
var match = AmbiguousFractionConnectorsRegex.Match(er.Text);
if (match.Success)
{
var beforeText = er.Text.Substring(0, match.Index).TrimEnd();
er.Length = beforeText.Length;
er.Text = beforeText;
er.Type = Constants.SYS_NUM;
er.Data = null;
}
}
}
return(ers);
}
public virtual List <ExtractResult> Extract(string source)
{
if (string.IsNullOrEmpty(source))
{
return(new List <ExtractResult>());
}
var result = new List <ExtractResult>();
var matchSource = new Dictionary <Match, TypeTag>();
var matched = new bool[source.Length];
var collections = Regexes.ToDictionary(o => o.Key.Matches(source), p => p.Value);
foreach (var collection in collections)
{
foreach (Match m in collection.Key)
{
// In ExperimentalMode, AmbiguousFraction like "30000 in 2009" needs to be skipped
if ((Options & NumberOptions.ExperimentalMode) != 0 && AmbiguousFractionConnectorsRegex.Match(m.Value).Success)
{
continue;
}
// In EnablePreview, cases like "last", "next" should not be skipped
if ((Options & NumberOptions.EnablePreview) == 0 && IsRelativeOrdinal(m.Value))
{
continue;
}
for (var j = 0; j < m.Length; j++)
{
matched[m.Index + j] = true;
}
// Fliter out cases like "first two", "last one"
// only support in English now
if (ExtractType.Contains(Constants.MODEL_ORDINAL) && RelativeOrdinalFilterRegex != null && RelativeOrdinalFilterRegex.IsMatch(source))
{
continue;
}
// Keep Source Data for extra information
matchSource.Add(m, collection.Value);
}
}
var last = -1;
for (var i = 0; i < source.Length; i++)
{
if (matched[i])
{
if (i + 1 == source.Length || !matched[i + 1])
{
var start = last + 1;
var length = i - last;
var substr = source.Substring(start, length);
if (matchSource.Keys.Any(o => o.Index == start && o.Length == length))
{
var type = matchSource.Where(p => p.Key.Index == start && p.Key.Length == length)
.Select(p => (p.Value.Priority, p.Value.Name)).Min().Item2;
// Extract negative numbers
if (NegativeNumberTermsRegex != null)
{
var match = NegativeNumberTermsRegex.Match(source.Substring(0, start));
if (match.Success)
{
start = match.Index;
length = length + match.Length;
substr = match.Value + substr;
}
}
var er = new ExtractResult
{
Start = start,
Length = length,
Text = substr,
Type = ExtractType,
Data = type,
};
// Add Metadata information for Ordinal
if (ExtractType.Contains(Constants.MODEL_ORDINAL))
{
er.Metadata = new Metadata();
if (IsRelativeOrdinal(substr))
{
er.Metadata.IsOrdinalRelative = true;
}
}
result.Add(er);
}
}
}
else
{
last = i;
}
}
result = FilterAmbiguity(result, source);
return(result);
}
// Fraction with InConnector may lead to some ambiguous cases like "more than 30000 in 2010"
// In ExperimentalMode, we will remove all FractionWithInConnector numbers to avoid such cases
private bool IsFractionWithInConnector(string numberStr)
{
return(AmbiguousFractionConnectorsRegex.Match(numberStr).Success);
}
public virtual List <ExtractResult> Extract(string source)
{
if (string.IsNullOrEmpty(source))
{
return(new List <ExtractResult>());
}
var result = new List <ExtractResult>();
var matchSource = new Dictionary <Match, TypeTag>();
var matched = new bool[source.Length];
var collections = Regexes.ToDictionary(o => o.Key.Matches(source), p => p.Value);
foreach (var collection in collections)
{
foreach (Match m in collection.Key)
{
// In ExperimentalMode, AmbiguousFraction like "30000 in 2009" needs to be skipped
if ((Options & NumberOptions.ExperimentalMode) != 0 && AmbiguousFractionConnectorsRegex.Match(m.Value).Success)
{
continue;
}
// If SuppressExtendedTypes is on, cases like "last", "next" should be skipped
if ((Options & NumberOptions.SuppressExtendedTypes) != 0 && m.Groups[Constants.RelativeOrdinalGroupName].Success)
{
continue;
}
// Matches containing separators 'in', 'out of' should be considered fractions only when numerator < denominator
if (m.Groups["ambiguousSeparator"].Success)
{
var numerator = m.Groups["numerator"];
var denominator = m.Groups["denominator"];
int num = ParseNumber(numerator);
int den = ParseNumber(denominator);
if (num > den)
{
continue;
}
}
for (var j = 0; j < m.Length; j++)
{
matched[m.Index + j] = true;
}
// Keep Source Data for extra information
matchSource.Add(m, collection.Value);
}
}
var last = -1;
for (var i = 0; i < source.Length; i++)
{
if (matched[i])
{
if (i + 1 == source.Length || !matched[i + 1])
{
var start = last + 1;
var length = i - last;
var substr = source.Substring(start, length);
if (matchSource.Keys.Any(o => o.Index == start && o.Length == length))
{
var(_, type, originalMatch) = matchSource.Where(p => p.Key.Index == start && p.Key.Length == length)
.Select(p => (p.Value.Priority, p.Value.Name, p.Key)).Min();
// Extract negative numbers
if (NegativeNumberTermsRegex != null)
{
var match = NegativeNumberTermsRegex.Match(source.Substring(0, start));
if (match.Success)
{
start = match.Index;
length += match.Length;
substr = match.Value + substr;
}
}
var er = new ExtractResult
{
Start = start,
Length = length,
Text = substr,
Type = ExtractType,
Data = type,
};
// Add Metadata information for Ordinal
if (ExtractType.Contains(Constants.MODEL_ORDINAL))
{
er.Metadata = new Metadata();
if ((Options & NumberOptions.SuppressExtendedTypes) == 0 &&
originalMatch.Groups[Constants.RelativeOrdinalGroupName].Success)
{
er.Metadata.IsOrdinalRelative = true;
}
}
result.Add(er);
}
}
}
else
{
last = i;
}
}
result = FilterAmbiguity(result, source);
return(result);
}