public void LoadFromGigaFidaFile(string fileName)
{
Utils.ThrowException(fileName == null ? new ArgumentNullException("fileName") : null);
Utils.ThrowException(!Utils.VerifyFileNameOpen(fileName) ? new ArgumentValueException("fileName") : null);
XmlTextReader xmlReader = null;
try
{
bool hasHeader = false;
mTaggedWords.Clear();
mTeiHeader = null;
xmlReader = new XmlTextReader(new FileStream(fileName, FileMode.Open));
while (xmlReader.Read())
{
if (xmlReader.NodeType == XmlNodeType.Element && xmlReader.Name == "teiHeader") // header
{
hasHeader = true;
Utils.XmlSkip(xmlReader, "teiHeader");
}
else if (xmlReader.NodeType == XmlNodeType.Element && xmlReader.Name == "p") // paragraph
{
ThreadHandler.AbortCheckpoint(); // TODO: do this at various appropriate places
xmlReader.Read();
Corpus aux = new Corpus();
aux.LoadFromTextSsjTokenizer(xmlReader.Value);
if (aux.TaggedWords.Count > 0)
{
foreach (TaggedWord word in aux.TaggedWords)
{
word.MoreInfo.RemoveEndOfParagraphFlag();
mTaggedWords.Add(word);
}
aux.TaggedWords.Last.MoreInfo.SetEndOfParagraphFlag();
}
}
}
xmlReader.Close();
if (hasHeader)
{
ReadTeiHeader(fileName);
}
}
catch
{
try { xmlReader.Close(); } catch { }
throw;
}
}
private static void Main(string[] args)
{
// initialize logger
mLogger.LocalLevel = Logger.Level.Debug;
mLogger.LocalOutputType = Logger.OutputType.Custom;
Logger.CustomOutput = new Logger.CustomOutputDelegate(delegate(string loggerName, Logger.Level level, string funcName, Exception e,
string message, object[] msgArgs) { Console.WriteLine(message, msgArgs); });
#endif
try
{
if (args.Length < 4)
{
OutputHelp();
}
else
{
string inputFolder = null, searchPattern = null, taggerModelFile = null, lemmatizerModelFile = null, outputFileOrFolder = null;
bool ssjTokenizer = false, searchSubfolders = false, verbose = false, overwrite = false;
if (ParseParams(args, ref verbose, ref inputFolder, ref searchPattern, ref taggerModelFile, ref lemmatizerModelFile,
ref outputFileOrFolder, ref ssjTokenizer, ref searchSubfolders, ref overwrite))
{
if (!verbose)
{
mLogger.LocalLevel = Logger.Level.Info;
mLogger.LocalProgressOutputType = Logger.ProgressOutputType.Off;
}
PartOfSpeechTagger tagger = new PartOfSpeechTagger();
tagger.LoadModels(taggerModelFile, lemmatizerModelFile);
mLogger.Debug(null, "Mapa z vhodnimi datotekami: {0}", inputFolder);
mLogger.Debug(null, "Iskalni vzorec: {0}", searchPattern);
foreach (FileInfo file in new DirectoryInfo(inputFolder).GetFiles(searchPattern,
searchSubfolders ? SearchOption.AllDirectories : SearchOption.TopDirectoryOnly))
{
string outputFileName = outputFileOrFolder;
if (searchPattern.Contains("*") || searchPattern.Contains("?")) // search pattern contains wildcards thus output is a folder
{
outputFileName = outputFileOrFolder.TrimEnd('\\') + "\\" +
file.Name.Substring(0, file.Name.Length - file.Extension.Length) + ".out" + file.Extension;
DirectoryInfo dirInfo = new FileInfo(outputFileName).Directory;
if (!dirInfo.Exists)
{
dirInfo.Create();
}
}
if (File.Exists(outputFileName) && !overwrite)
{
mLogger.Debug(null, "Datoteka {0} že obstaja. Pripadajoča vhodna datoteka ni bila ponovno označena.", outputFileName);
continue;
}
mLogger.Debug(null, "Nalagam {0} ...", file.FullName);
Corpus corpus;
bool xmlMode = false;
try
{
corpus = new Corpus();
corpus.LoadFromXmlFile(file.FullName, /*tagLen=*/ -1);
if (corpus.TaggedWords.Count > 0)
{
xmlMode = true;
}
else
{
corpus.LoadFromGigaFidaFile(file.FullName);
}
}
catch (ThreadHandler.AbortedByUserException)
{
throw;
}
catch (Exception e)
{
corpus = new Corpus();
string content = File.ReadAllText(file.FullName);
if (IsXmlTei(content))
{
mLogger.Debug(null, "*** Opozorilo: Datoteka z besedilom vsebuje značke XML-TEI, vendar nima pravilne oblike. Podrobnosti: {0}", e.Message);
}
if (ssjTokenizer)
{
corpus.LoadFromTextSsjTokenizer(content);
}
else
{
corpus.LoadFromText(content);
}
}
int knownWordsCorrect = 0;
int knownWordsPosCorrect = 0;
int knownWords = 0;
int unknownWordsCorrect = 0;
int unknownWordsPosCorrect = 0;
int unknownWords = 0;
int eosCount = 0;
int eosCorrect = 0;
int lemmaCorrect = 0;
int lemmaCorrectLowercase = 0;
int lemmaWords = 0;
int knownWordsCorrectNoPunct = 0;
int knownWordsPosCorrectNoPunct = 0;
int knownWordsNoPunct = 0;
int unknownWordsCorrectNoPunct = 0;
int unknownWordsPosCorrectNoPunct = 0;
int unknownWordsNoPunct = 0;
string[] goldTags = new string[corpus.TaggedWords.Count];
for (int i = 0; i < corpus.TaggedWords.Count; i++)
{
goldTags[i] = corpus.TaggedWords[i].Tag;
}
tagger.Tag(corpus, out lemmaCorrect, out lemmaCorrectLowercase, out lemmaWords, xmlMode);
mLogger.Debug(null, "Zapisujem označeno besedilo v datoteko {0} ...", outputFileName);
StreamWriter writer = new StreamWriter(outputFileName);
writer.Write(corpus.ToString(xmlMode || ssjTokenizer ? "XML-MI" : "XML"));
writer.Close();
mLogger.Debug(null, "Končano.");
if (xmlMode)
{
for (int i = 0; i < corpus.TaggedWords.Count; i++)
{
string wordLower = corpus.TaggedWords[i].WordLower;
string tag = corpus.TaggedWords[i].Tag;
bool isKnown = tagger.IsKnownWord(wordLower);
if (!PartOfSpeechTagger.mNonWordRegex.Match(corpus.TaggedWords[i].Word).Success)
{
if (tag == goldTags[i])
{
if (isKnown)
{
knownWordsCorrectNoPunct++;
}
else
{
unknownWordsCorrectNoPunct++;
}
}
if (goldTags[i] != null && tag[0] == goldTags[i][0])
{
if (isKnown)
{
knownWordsPosCorrectNoPunct++;
}
else
{
unknownWordsPosCorrectNoPunct++;
}
}
if (isKnown)
{
knownWordsNoPunct++;
}
else
{
unknownWordsNoPunct++;
}
}
if (tag == goldTags[i])
{
if (isKnown)
{
knownWordsCorrect++;
}
else
{
unknownWordsCorrect++;
}
}
if (goldTags[i] != null && tag[0] == goldTags[i][0])
{
if (isKnown)
{
knownWordsPosCorrect++;
}
else
{
unknownWordsPosCorrect++;
}
}
if (isKnown)
{
knownWords++;
}
else
{
unknownWords++;
}
if (corpus.TaggedWords[i].MoreInfo.EndOfSentence)
{
eosCount++;
if (tag.EndsWith("<eos>"))
{
eosCorrect++;
}
}
}
int allWords = knownWords + unknownWords;
int allWordsCorrect = knownWordsCorrect + unknownWordsCorrect;
int allWordsPosCorrect = knownWordsPosCorrect + unknownWordsPosCorrect;
int allWordsNoPunct = knownWordsNoPunct + unknownWordsNoPunct;
int allWordsCorrectNoPunct = knownWordsCorrectNoPunct + unknownWordsCorrectNoPunct;
int allWordsPosCorrectNoPunct = knownWordsPosCorrectNoPunct + unknownWordsPosCorrectNoPunct;
mLogger.Info(null, "Točnost na znanih besedah: ................... {2:0.00}% ({0} / {1})", knownWordsCorrect, knownWords,
(double)knownWordsCorrect / (double)knownWords * 100.0);
mLogger.Info(null, "Točnost na neznanih besedah: ................. {2:0.00}% ({0} / {1})", unknownWordsCorrect, unknownWords,
(double)unknownWordsCorrect / (double)unknownWords * 100.0);
mLogger.Info(null, "Skupna točnost: .............................. {2:0.00}% ({0} / {1})", allWordsCorrect, allWords,
(double)allWordsCorrect / (double)allWords * 100.0);
mLogger.Info(null, "Točnost na znanih besedah (POS): ............. {2:0.00}% ({0} / {1})", knownWordsPosCorrect, knownWords,
(double)knownWordsPosCorrect / (double)knownWords * 100.0);
mLogger.Info(null, "Točnost na neznanih besedah (POS): ........... {2:0.00}% ({0} / {1})", unknownWordsPosCorrect, unknownWords,
(double)unknownWordsPosCorrect / (double)unknownWords * 100.0);
mLogger.Info(null, "Skupna točnost (POS): ........................ {2:0.00}% ({0} / {1})", allWordsPosCorrect, allWords,
(double)allWordsPosCorrect / (double)allWords * 100.0);
mLogger.Info(null, "Točnost na znanih besedah (brez ločil): ...... {2:0.00}% ({0} / {1})", knownWordsCorrectNoPunct, knownWordsNoPunct,
(double)knownWordsCorrectNoPunct / (double)knownWordsNoPunct * 100.0);
mLogger.Info(null, "Točnost na neznanih besedah (brez ločil): .... {2:0.00}% ({0} / {1})", unknownWordsCorrectNoPunct, unknownWordsNoPunct,
(double)unknownWordsCorrectNoPunct / (double)unknownWordsNoPunct * 100.0);
mLogger.Info(null, "Skupna točnost (brez ločil): ................. {2:0.00}% ({0} / {1})", allWordsCorrectNoPunct, allWordsNoPunct,
(double)allWordsCorrectNoPunct / (double)allWordsNoPunct * 100.0);
mLogger.Info(null, "Točnost na znanih besedah (POS, brez ločil): {2:0.00}% ({0} / {1})", knownWordsPosCorrectNoPunct, knownWordsNoPunct,
(double)knownWordsPosCorrectNoPunct / (double)knownWordsNoPunct * 100.0);
mLogger.Info(null, "Točnost na neznanih besedah (POS, brez ločil): {2:0.00}% ({0} / {1})", unknownWordsPosCorrectNoPunct, unknownWordsNoPunct,
(double)unknownWordsPosCorrectNoPunct / (double)unknownWordsNoPunct * 100.0);
mLogger.Info(null, "Skupna točnost (POS, brez ločil): ............ {2:0.00}% ({0} / {1})", allWordsPosCorrectNoPunct, allWordsNoPunct,
(double)allWordsPosCorrectNoPunct / (double)allWordsNoPunct * 100.0);
if (lemmatizerModelFile != null)
{
mLogger.Info(null, "Točnost lematizacije (brez ločil): ........... {2:0.00}% ({0} / {1})", lemmaCorrect, lemmaWords,
(double)lemmaCorrect / (double)lemmaWords * 100.0);
mLogger.Info(null, "Točnost lematizacije (male črke, brez ločil): {2:0.00}% ({0} / {1})", lemmaCorrectLowercase, lemmaWords,
(double)lemmaCorrectLowercase / (double)lemmaWords * 100.0);
}
mLogger.Info(null, "Točnost detekcije konca stavka: .............. {2:0.00}% ({0} / {1})", eosCorrect, eosCount,
(double)eosCorrect / (double)eosCount * 100.0);
}
}
}
}
}
catch (Exception e)
{
mLogger.Info(null, "");
mLogger.Info(null, "*** Nepričakovana napaka. Podrobnosti: {0}", e);
}
}