public T GetRandomEntity <T>(string language) where T : class, IMarcellEntity
{
T result;
do
{
byte[] rndBytes = new byte[4];
m_randomGenerator.GetBytes(rndBytes);
double dblValue = BitConverter.ToUInt32(rndBytes, 0);
double factor = dblValue / uint.MaxValue;
IndexReader selectedReader;
if (typeof(T) == typeof(Document))
{
selectedReader = m_documentReader[language];
}
else if (typeof(T) == typeof(Section))
{
selectedReader = m_sectionReader[language];
}
else if (typeof(T) == typeof(Paragraph))
{
selectedReader = m_paragraphReader[language];
}
else if (typeof(T) == typeof(Sentence))
{
selectedReader = m_sentenceReader[language];
}
else
{
throw new InvalidOperationException("Unsupported request received!");
}
int docNo = (int)(factor * (selectedReader.MaxDoc - 1));
LuceneNet.Documents.Document outputDoc = selectedReader.Document(docNo);
if (typeof(T) == typeof(Document))
{
result = outputDoc.ToDocument() as T;
}
else if (typeof(T) == typeof(Section))
{
result = outputDoc.ToSection() as T;
}
else if (typeof(T) == typeof(Paragraph))
{
result = outputDoc.ToParagraph() as T;
}
else if (typeof(T) == typeof(Sentence))
{
result = outputDoc.ToSentence() as T;
}
else
{
throw new InvalidOperationException("Unsupported entity type requested!");
}
} while (result.TokenCount < 5 || result.RecognitionQuality < 0.10); //Only use sufficiently large text chunks with enough recognized tokens for training
return(result);
}
