public void WikipediaCases(string input, string output)
{
var result = StateMachineTokenizer.Tokenize(input.ToCharArray(), lowerCase: false);
var actual = new string(result.ToArray());
Assert.Equal(output, actual);
}
public void SpecialCase(string input, string output, bool lowerCase)
{
var result = StateMachineTokenizer.Tokenize(input.ToCharArray(), lowerCase);
var actual = new string(result.ToArray());
Assert.Equal(output, actual);
}
private static IEnumerable <int> Process(IList <char> text)
{
var cleaned = WikitextCleaner.Clean(text);
var tokenized = StateMachineTokenizer.Tokenize(cleaned, lowerCase: true);
return(NaiveTokenizer.Tokenize(tokenized)
.Select(ft => TextHasher.CalculateHashCode(tokenized, ft.From, ft.To)));
}
static void TokenizeWikipedia()
{
var outputWikipediaPath = tokenizedPath;
PrepareOutputDirectory(outputWikipediaPath);
var reader = new CorpusZipReader <IList <char> >(cleanedPath, charDataSerializer);
var writer = new CorpusZipWriter <IList <char> >(outputWikipediaPath, charDataSerializer);
var tokenizer = new CorpusTransformer <IList <char>, IList <char> >(t => StateMachineTokenizer.Tokenize(t, lowerCase: true));
tokenizer.Transform(reader, writer);
}
