@Test
public void testFractions() {
String[] sample = {"5-1/4 plus 2 3/16 = 7\u00A07/16 in the U.S.S.R. Why not?"};
String[][] tokenizedNormal = {
{
"5-1/4",
"plus",
"2\u00A03/16",
"=",
"7\u00A07/16",
"in",
"the",
"U.S.S.R.",
".",
"Why",
"not",
"?"
}
};
String[][] tokenizedStrict = {
{
"5-1/4", "plus", "2", "3/16", "=", "7", "7/16", "in", "the", "U.S.S.R", ".", "Why", "not",
"?"
}
};
TokenizerFactory<CoreLabel> tokFactoryNormal = PTBTokenizer.coreLabelFactory();
TokenizerFactory<CoreLabel> tokFactoryStrict = PTBTokenizer.coreLabelFactory("strictTreebank3");
runOnTwoArrays(tokFactoryNormal, sample, tokenizedNormal);
runOnTwoArrays(tokFactoryStrict, sample, tokenizedStrict);
}
@Test
public void testPTBTokenizerSGML() {
TokenizerFactory<CoreLabel> tokFactory = PTBTokenizer.coreLabelFactory();
runOnTwoArrays(tokFactory, sgmlInputs, sgmlGold);
}
@Test
public void testPTBTokenizerTokenizeSplitHyphens() {
TokenizerFactory<CoreLabel> tokFactory = PTBTokenizer.coreLabelFactory("splitHyphenated=true");
runOnTwoArrays(tokFactory, ptbInputs, ptbGoldSplitHyphenated);
}
@Test
public void testPTBTokenizerTokenizePerLineSGML() {
TokenizerFactory<CoreLabel> tokFactory = PTBTokenizer.coreLabelFactory("tokenizePerLine=true");
runOnTwoArrays(tokFactory, sgmlInputs, sgmlPerLineGold);
}