concordia-library/concordia/sentence_tokenizer.cpp

#include "concordia/sentence_tokenizer.hpp"
#include "concordia/token_annotation.hpp"

#include <boost/foreach.hpp>
#include <fstream>
#include <sstream>
#include <iostream>
#include <boost/algorithm/string.hpp>

SentenceTokenizer::SentenceTokenizer(
                        boost::shared_ptr<ConcordiaConfig> config)
                                         throw(ConcordiaException) {
    _createNeRules(config->getNamedEntitiesFilePath());
    _createHtmlTagsRule(config->getHtmlTagsFilePath());
    _stopWordsEnabled = config->isStopWordsEnabled();
    if (_stopWordsEnabled) {
        _stopWords = _getMultipleRegexRule(
                                  config->getStopWordsFilePath(),
                                  TokenAnnotation::STOP_WORD,
                                  "", true);
    }
}

SentenceTokenizer::~SentenceTokenizer() {
}

TokenizedSentence SentenceTokenizer::tokenize(const std::string & sentence,
                                              bool byWhitespace) {
    TokenizedSentence result(sentence);

    if (byWhitespace) {
        result.toLowerCase();

        boost::shared_ptr<RegexRule> whitespaceRule(
                            new RegexRule("\\S+",
                                          TokenAnnotation::WORD, ""));
        whitespaceRule->apply(result);
    } else {
        _htmlTags->apply(result);

        BOOST_FOREACH(RegexRule & neRule, _namedEntities) {
            neRule.apply(result);
        }

        result.toLowerCase();

        if (_stopWordsEnabled) {
            _stopWords->apply(result);
        }

        boost::shared_ptr<RegexRule> wordsWithNumbersRule(
                                    new RegexRule("[0-9]+\\-(\\p{L}|[0-9])+",
                                                  TokenAnnotation::WORD, ""));
                wordsWithNumbersRule->apply(result);

        boost::shared_ptr<RegexRule> numbersRule(
                            new RegexRule("\\b[0-9]+([\\.\\,][0-9]+)?\\b", TokenAnnotation::NE, "ne_number"));
        numbersRule->apply(result);

        boost::shared_ptr<RegexRule> wordsRule(
                            new RegexRule("(\\p{L}|[0-9])(\\p{L}|[0-9]|'|\\-)*(\\p{L}|[0-9])",
                                          TokenAnnotation::WORD, ""));
        wordsRule->apply(result);

        boost::shared_ptr<RegexRule> singleLetterWordsRule(
                            new RegexRule("\\p{L}", TokenAnnotation::WORD, ""));
        singleLetterWordsRule->apply(result);


    }

    return result;
}

void SentenceTokenizer::_createNeRules(std::string & namedEntitiesPath) {
    if (boost::filesystem::exists(namedEntitiesPath)) {
        std::string line;
        std::ifstream neFile(namedEntitiesPath.c_str());
        if (neFile.is_open()) {
            int lineCounter = 0;
            while (getline(neFile, line)) {
                lineCounter++;
                boost::shared_ptr<std::vector<std::string> >
                                 tokenTexts(new std::vector<std::string>());
                boost::split(*tokenTexts, line, boost::is_any_of(" "),
                             boost::token_compress_on);
                if (tokenTexts->size() != 2) {
                    std::stringstream ss;
                    ss << "Invalid line: " << lineCounter
                       << " in NE file: " << namedEntitiesPath;
                    throw ConcordiaException(ss.str());
                } else {
                    _namedEntities.push_back(RegexRule(
                                tokenTexts->at(0),
                                TokenAnnotation::NE,
                                tokenTexts->at(1)));
                }
            }
            neFile.close();
        } else {
            throw ConcordiaException("Unable to read named entities file.");
        }
    } else {
        throw ConcordiaException("No named entities file.");
    }
}

void SentenceTokenizer::_createHtmlTagsRule(std::string & htmlTagsPath) {
    std::string tagsExpression = "<\\/?(";
    if (boost::filesystem::exists(htmlTagsPath)) {
        std::string line;
        std::ifstream tagsFile(htmlTagsPath.c_str());
        if (tagsFile.is_open()) {
            while (getline(tagsFile, line)) {
                tagsExpression += line +"|";
            }
           tagsFile.close();
        } else {
            throw ConcordiaException("Unable to read html tags file.");
        }
    } else {
        throw ConcordiaException("No html tags file.");
    }
    tagsExpression = tagsExpression.substr(0, tagsExpression.size()-1);
    tagsExpression += ").*?>";
    _htmlTags = boost::shared_ptr<RegexRule>(
                        new RegexRule(tagsExpression,
                                TokenAnnotation::HTML_TAG, "", false));
}

boost::shared_ptr<RegexRule>
        SentenceTokenizer::_getMultipleRegexRule(
            std::string filePath,
            char annotationType,
            std::string value,
            bool wholeWord) {
    std::string expression = "(";
    if (boost::filesystem::exists(filePath)) {
        std::string line;
        std::ifstream ruleFile(filePath.c_str());
        if (ruleFile.is_open()) {
            while (getline(ruleFile, line)) {
                if (wholeWord) {
                    expression += "\\b";
                }
                expression += line;
                if (wholeWord) {
                    expression += "\\b";
                }
                expression += "|";
            }
           ruleFile.close();
        } else {
            throw ConcordiaException("Unable to read file: "+filePath);
        }
    } else {
        throw ConcordiaException("No "+filePath+" file.");
    }
    expression = expression.substr(0, expression.size()-1);
    expression += ")";
    return boost::shared_ptr<RegexRule>(
                      new RegexRule(expression, annotationType, value, false));
}