public static Dictionary<string, List<StringProbabEntry>> GetTransliterations(Dictionary<string, int> lowerCasedTerms, MPAlignerConfigurationTranslEntry translEntry, string mosesPath, string tempFilePath, int threadCount)
{
Dictionary<string, List<StringProbabEntry>> res = new Dictionary<string, List<StringProbabEntry>> ();
if (translEntry == null || lowerCasedTerms == null || lowerCasedTerms.Count < 1 || string.IsNullOrWhiteSpace (mosesPath) || string.IsNullOrWhiteSpace (tempFilePath)) {
return res;
}
string langKey = translEntry != null ? ((translEntry.srcLang != null ? translEntry.srcLang : "") + "_" + (translEntry.trgLang != null ? translEntry.trgLang : "")) : "";
Log.Write ("Starting transliteration of " + lowerCasedTerms.Count.ToString () + " tokens.", LogLevelType.LIMITED_OUTPUT);
int idx = 0;
List<List<string>> lowerCasedTermDictList = new List<List<string>> (threadCount);
for (int i=0; i<threadCount; i++) {
lowerCasedTermDictList.Add (new List<string> ());
}
foreach (string term in lowerCasedTerms.Keys) {
lowerCasedTermDictList [idx % threadCount].Add (term);
idx++;
}
string directory = Path.GetDirectoryName (mosesPath);
List<Process> processes = new List<Process> ();
for (int i=0; i<lowerCasedTermDictList.Count; i++) {
if (lowerCasedTermDictList [i].Count > 0) {
try {
string tmpFile = tempFilePath + i.ToString () + ".tmp";
WriteWordsForTransliteration (lowerCasedTermDictList [i], tmpFile);
ProcessStartInfo myProcessStartInfo = new ProcessStartInfo (mosesPath);
myProcessStartInfo.UseShellExecute = false;
myProcessStartInfo.WorkingDirectory = directory;
myProcessStartInfo.FileName = mosesPath;
myProcessStartInfo.CreateNoWindow = true;
myProcessStartInfo.RedirectStandardOutput = true;
myProcessStartInfo.RedirectStandardError = true;
StringBuilder sb = new StringBuilder ();
sb.Append (" -f ");
sb.Append ("\"" + translEntry.mosesIniPath + "\" ");
sb.Append (" -i ");
sb.Append ("\"" + tmpFile + "\" ");
sb.Append (" -n-best-list ");
sb.Append ("\"" + tmpFile + ".n_best\" " + translEntry.nBest.ToString ());
myProcessStartInfo.Arguments = sb.ToString ();
processes.Add (new Process ());
processes [processes.Count - 1].StartInfo = myProcessStartInfo;
bool started = processes [processes.Count - 1].Start ();
processes [processes.Count - 1].ErrorDataReceived += p_ErrorDataReceived;
processes [processes.Count - 1].OutputDataReceived += p_OutputDataReceived;
processes [processes.Count - 1].BeginOutputReadLine ();
processes [processes.Count - 1].BeginErrorReadLine ();
} catch {
}
}
}
for (int i=0; i<processes.Count; i++) {
processes [i].WaitForExit ();
processes [i].Close ();
processes [i].Dispose ();
}
processes.Clear ();
for (int i=0; i<lowerCasedTermDictList.Count; i++) {
if (lowerCasedTermDictList[i].Count > 0) {
string tmpFile = tempFilePath + i.ToString () + ".tmp";
if (File.Exists (tmpFile + ".n_best")) {
NumberFormatInfo nfi = new NumberFormatInfo ();
nfi.CurrencyDecimalSeparator = ".";
nfi.NumberDecimalSeparator = ".";
nfi.PercentDecimalSeparator = ".";
Dictionary<string,Dictionary<string,bool>> existingTranslits = new Dictionary<string, Dictionary<string,bool>> ();
StreamReader sr = new StreamReader (tmpFile + ".n_best", Encoding.UTF8);
string[] sep = {"|||"};
while (!sr.EndOfStream) {
string line = sr.ReadLine ();
string[] dataArr = line.Split (sep, StringSplitOptions.RemoveEmptyEntries);
if (dataArr.Length == 4) {
try {
string idStr = dataArr [0];
idStr = idStr.Trim ();
int id = Convert.ToInt32 (idStr);
string word = dataArr [1];
StringProbabEntry spe = new StringProbabEntry ();
spe.str = word.Trim ().Replace (" ", "");
string probabStr = dataArr [3];
probabStr = probabStr.Trim ().Replace (',', '.');
spe.probab = Math.Exp (Convert.ToDouble (probabStr, nfi));
if (spe.probab>1) spe.probab = 1;
if (id < lowerCasedTermDictList[i].Count) {
string term = lowerCasedTermDictList[i][id];
double min = Math.Min (spe.str.Length, term.Length);
double max = Math.Max (spe.str.Length, term.Length);
double lenDiff = min / max;
//Log.Write(term+" "+word+" "+lenDiff.ToString()+" "+spe.probab.ToString(),LogLevelType.ERROR);
if (lenDiff >= translEntry.maxLenDiff) {
if (!existingTranslits.ContainsKey (term))
existingTranslits.Add (term, new Dictionary<string,bool> ());
if (!res.ContainsKey (term))
res.Add (term, new List<StringProbabEntry> ());
if (!translitTemp[langKey].ContainsKey(term)) translitTemp[langKey].Add(term, new List<StringProbabEntry>());
if (!existingTranslits [term].ContainsKey (spe.str) && spe.probab >= translEntry.threshold) {
spe.probab = translEntry.translitBf.Get (spe.probab);
existingTranslits [term].Add (spe.str, true);
res [term].Add (spe);
translitTemp[langKey][term].Add(spe);
}
}
}
} catch {
}
}
}
}
try {
File.Delete (tmpFile + ".n_best");
File.Delete (tmpFile);
} catch {
}
}
}
GC.Collect();
GC.WaitForPendingFinalizers();
return res;
}
private static void CopyTranslits(Dictionary<string, List<StringProbabEntry>> fromDict, Dictionary<string, List<StringProbabEntry>> toDict, MPAlignerConfigurationTranslEntry translEntry)
{
string langKey = translEntry != null ? ((translEntry.srcLang != null ? translEntry.srcLang : "") + "_" + (translEntry.trgLang != null ? translEntry.trgLang : "")) : "";
if (fromDict != null) {
if (toDict == null)
{
toDict = new Dictionary<string, List<StringProbabEntry>>();
}
foreach(string term in fromDict.Keys)
{
if (!toDict.ContainsKey(term))
{
toDict.Add(term, fromDict[term]);
if (!translitTemp[langKey].ContainsKey(term)) translitTemp[langKey].Add(term, new List<StringProbabEntry>());
translitTemp[langKey][term]=fromDict[term];
}
}
}
}
public static List<ProcessedTermEntry> ProcessTermsList(List<string> terms, Dictionary<string, Dictionary<string, double>> srcToTrgDict, string lang, MPAlignerConfigurationTranslEntry translitEntry, string mosesPath, string tempFilePath, int threadCount = 1 , bool stemWords = false)
{
List<ProcessedTermEntry> res = new List<ProcessedTermEntry>(1000);
Dictionary<string, int> lowercasedWordDict = new Dictionary<string, int>(1000);
if (terms!=null)
{
string langKey = translitEntry!=null?((translitEntry.srcLang!=null?translitEntry.srcLang:"")+"_"+(translitEntry.trgLang!=null?translitEntry.trgLang:"")):lang;
if (!translitTemp.ContainsKey(langKey)) translitTemp.Add(langKey,new Dictionary<string, List<StringProbabEntry>>());
foreach(string surfaceForm in terms)
{
string lowerCase = surfaceForm.ToLower();
ProcessedTermEntry pte = new ProcessedTermEntry();
pte.surfaceForm = surfaceForm;
pte.lowercaceForm = lowerCase;
pte.surfaceFormWords = new List<string>(whitespaceRegex.Split(surfaceForm));
string[] lowerCaseWordArr = whitespaceRegex.Split(lowerCase);
pte.lowercaseWords.InsertRange(0,lowerCaseWordArr);
foreach(string word in lowerCaseWordArr)
{
pte.len+=word.Length;
if (!lowercasedWordDict.ContainsKey(word) && !translitTemp[langKey].ContainsKey(word))
{
lowercasedWordDict.Add(word,0);
}
string stem = null;
if (stemWords)
{
stem = LightweightStemmer.Stem(word,lang);
}
//if (lang !="en")
//{
pte.simpleTransliteration.Add(SimpleCharacterTransliteration.Transliterate(word));
//}
//else
//{
// pte.simpleTransliteration = pte.lowercaseWords;
//}
if (srcToTrgDict!=null)
{
List<StringProbabEntry> currList = new List<StringProbabEntry>();
if (stemWords)
{
if (srcToTrgDict.ContainsKey(stem))
{
foreach(string trgStem in srcToTrgDict[stem].Keys)
{
StringProbabEntry spe = new StringProbabEntry();
spe.str=trgStem;
spe.probab = srcToTrgDict[stem][trgStem];
currList.Add(spe);
}
}
}
else
{
if (srcToTrgDict.ContainsKey(word))
{
foreach(string trgWord in srcToTrgDict[word].Keys)
{
StringProbabEntry spe = new StringProbabEntry();
spe.str=trgWord;
spe.probab = srcToTrgDict[word][trgWord];
currList.Add(spe);
}
}
}
pte.translationList.Add(currList);
}
}
res.Add(pte);
}
Dictionary<string, List<StringProbabEntry>> translitDict = new Dictionary<string, List<StringProbabEntry>>();
//if (threadCount<2)
//{
translitDict = GetTransliterations(lowercasedWordDict,translitEntry, mosesPath, tempFilePath, threadCount);
//This is not nice, however necessary due to the multi-threaded execution - the temp list is not updated in the single-thread scenario
// Dictionary<string, List<StringProbabEntry>> tmp = new Dictionary<string, List<StringProbabEntry>>();
// CopyTranslits(translitDict,tmp, translitEntry);
//}
//else
//{
// translitDict = GetTransliterationsMultiThreaded(lowercasedWordDict,translitEntry, mosesPath, tempFilePath, threadCount);
//}
for(int i=0; i<res.Count;i++)
{
foreach(string word in res[i].lowercaseWords)
{
if (translitDict.ContainsKey(word))
{
res[i].transliterationList.Add(translitDict[word]);
}
else if (translitTemp.ContainsKey(langKey)&&translitTemp[langKey].ContainsKey(word))
{
res[i].transliterationList.Add(translitTemp[langKey][word]);
}
else
{
res[i].transliterationList.Add(new List<StringProbabEntry>());
}
}
}
//We add a simple data amount threshold in order not to overflow the memory ...
if (translitTemp[langKey].Count>=50000)
{
translitTemp[langKey].Clear();
GC.Collect();
}
}
return res;
}
public static Dictionary<string, ProcessedTermEntry> ProcessTerms(Dictionary<string, SimpleTermEntry> terms, Dictionary<string,Dictionary<string,double>> srcToTrgDict, string lang, MPAlignerConfigurationTranslEntry translitEntry, string mosesPath, string tempFilePath, int threadCount = 1 , bool stemWords = false)
{
Dictionary<string, ProcessedTermEntry> res = new Dictionary<string, ProcessedTermEntry>(1000);
Dictionary<string, int> lowercasedWordDict = new Dictionary<string, int>(1000);
if (terms!=null)
{
Log.Write ("Starting pre-processing of "+terms.Count.ToString()+" "+ lang +" terms.",LogLevelType.LIMITED_OUTPUT);
string langKey = translitEntry!=null?((translitEntry.srcLang!=null?translitEntry.srcLang:"")+"_"+(translitEntry.trgLang!=null?translitEntry.trgLang:"")):lang;
if (!translitTemp.ContainsKey(langKey)) translitTemp.Add(langKey,new Dictionary<string, List<StringProbabEntry>>());
foreach(string lowerCase in terms.Keys)
{
string surfaceForm = terms[lowerCase].term;
if (!res.ContainsKey(lowerCase)) //TODO: Nothing to do, but be aware that here we allow only the first capitalization of a surface form ... we will ignore other capitalizations.
{
ProcessedTermEntry pte = new ProcessedTermEntry();
pte.surfaceForm = surfaceForm;
pte.concordance = !string.IsNullOrWhiteSpace (terms [lowerCase].conc) ? terms [lowerCase].conc : "";
pte.normMsdSeq = !string.IsNullOrWhiteSpace (terms [lowerCase].normMsdSeq) ? new List<string> (whitespaceRegex.Split (terms [lowerCase].normMsdSeq)) : new List<string> ();
pte.normSeq = !string.IsNullOrWhiteSpace (terms [lowerCase].normSeq) ? new List<string>(whitespaceRegex.Split(terms[lowerCase].normSeq)):new List<string>();
pte.lowercaceForm = lowerCase;
pte.surfaceFormWords = !string.IsNullOrWhiteSpace (surfaceForm) ? new List<string> (whitespaceRegex.Split (surfaceForm)) : new List<string> ();
string[] lowerCaseWordArr = !string.IsNullOrWhiteSpace (lowerCase) ? whitespaceRegex.Split (lowerCase) : null;
if (lowerCaseWordArr!=null)
{
pte.lowercaseWords.InsertRange(0,lowerCaseWordArr);
}
if (!string.IsNullOrWhiteSpace(terms[lowerCase].lemmaSeq))
{
pte.lemmaSeq = new List<string>(whitespaceRegex.Split(terms[lowerCase].lemmaSeq));
}
else
{
pte.lemmaSeq = new List<string>();
for (int i=0;i<pte.lowercaseWords.Count;i++){pte.lemmaSeq.Add("");}
}
if (!string.IsNullOrWhiteSpace(terms[lowerCase].msdSeq))
{
pte.msdSeq = new List<string>(whitespaceRegex.Split(terms[lowerCase].msdSeq));
}
else
{
pte.msdSeq = new List<string>();
for (int i=0;i<pte.lowercaseWords.Count;i++){pte.msdSeq.Add("");}
}
foreach(string word in lowerCaseWordArr)
{
pte.len+=word.Length;
if (!lowercasedWordDict.ContainsKey(word) && !translitTemp[langKey].ContainsKey(word))
{
lowercasedWordDict.Add(word,0);
}
string stem = null;
if (stemWords)
{
stem = LightweightStemmer.Stem(word,lang);
}
//if (lang !="en")
//{
pte.simpleTransliteration.Add(SimpleCharacterTransliteration.Transliterate(word));
//}
//else
//{
// pte.simpleTransliteration = pte.lowercaseWords;
//}
if (srcToTrgDict!=null)
{
List<StringProbabEntry> currList = new List<StringProbabEntry>();
if (stemWords)
{
if (srcToTrgDict.ContainsKey(stem))
{
foreach(string trgStem in srcToTrgDict[stem].Keys)
{
StringProbabEntry spe = new StringProbabEntry();
spe.str=trgStem;
spe.probab = srcToTrgDict[stem][trgStem];
currList.Add(spe);
}
}
}
else
{
if (srcToTrgDict.ContainsKey(word))
{
foreach(string trgWord in srcToTrgDict[word].Keys)
{
StringProbabEntry spe = new StringProbabEntry();
spe.str=trgWord;
spe.probab = srcToTrgDict[word][trgWord];
currList.Add(spe);
}
}
}
pte.translationList.Add(currList);
}
}
res.Add(lowerCase,pte);
}
}
Dictionary<string, List<StringProbabEntry>> translitDict = new Dictionary<string, List<StringProbabEntry>>();
//if (threadCount<2)
//{
translitDict = GetTransliterations(lowercasedWordDict,translitEntry, mosesPath, tempFilePath,threadCount);
//Dictionary<string, List<StringProbabEntry>> tmp = new Dictionary<string, List<StringProbabEntry>>();
//CopyTranslits(translitDict,tmp, translitEntry);
//}
//else
//{
// translitDict = GetTransliterationsMultiThreaded(lowercasedWordDict,translitEntry, mosesPath, tempFilePath, threadCount);
//}
foreach (string lowerCase in res.Keys)
{
foreach(string word in res[lowerCase].lowercaseWords)
{
if (translitDict.ContainsKey(word))
{
res[lowerCase].transliterationList.Add(translitDict[word]);
}
else if (translitTemp.ContainsKey(langKey)&&translitTemp[langKey].ContainsKey(word))
{
res[lowerCase].transliterationList.Add(translitTemp[langKey][word]);
}
else
{
res[lowerCase].transliterationList.Add(new List<StringProbabEntry>());
}
}
}
//We add a simple data amount threshold in order not to overflow the memory ...
if (translitTemp[langKey].Count>=25000)
{
translitTemp[langKey].Clear();
GC.Collect();
}
}
return res;
}
public static bool GetTranslitConfig(MPAlignerConfiguration configuration, string srcLang, string trgLang, out MPAlignerConfigurationTranslEntry srcTranslitConf, out MPAlignerConfigurationTranslEntry trgTranslitConf, out MPAlignerConfigurationTranslEntry srcToTrgTranslitConf, out MPAlignerConfigurationTranslEntry trgToSrcTranslitConf)
{
Log.Write ("Searching for transliteration configurations.",LogLevelType.LIMITED_OUTPUT,configuration);
srcTranslitConf = null;
trgTranslitConf = null;
srcToTrgTranslitConf = null;
trgToSrcTranslitConf = null;
string srcLangKey = srcLang+"_en";
string trgLangKey = trgLang+"_en";
string langKey = srcLang+"_"+trgLang;
string langKey2 = trgLang+"_"+srcLang;
//Define transliteration directions and whether or not to use EN as interlingua.
if (configuration.forceEnTranslitInterlingua && configuration.translConfEntryDict.ContainsKey(srcLangKey)&&configuration.translConfEntryDict.ContainsKey(trgLangKey) && configuration.translConfEntryDict[srcLangKey].use && configuration.translConfEntryDict[trgLangKey].use)
{
srcTranslitConf= configuration.translConfEntryDict[srcLangKey];
trgTranslitConf= configuration.translConfEntryDict[trgLangKey];
Log.Write ("EN interlingua transliteration loaded for language "+srcLang+": "+ configuration.translConfEntryDict[srcLangKey].mosesIniPath, LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("EN interlingua transliteration loaded for language "+trgLang+": "+ configuration.translConfEntryDict[trgLangKey].mosesIniPath, LogLevelType.LIMITED_OUTPUT,configuration);
return true;
}else if (configuration.forceEnTranslitInterlingua){
Log.Write ("Cannot force EN interlingua transliteration for the pair "+langKey+" as at least one of the interlingua transliteration configurations is disabled or missing!",LogLevelType.WARNING,configuration);
Log.Write ("Will try falling back to direct transliteration without the EN interlingua.",LogLevelType.WARNING,configuration);
}
//If EN interlingua should not be used or one of the interlingua transliteration configurations is missing, try the direct transliteration.
bool foundAtLeastOne = false;
if (configuration.translConfEntryDict.ContainsKey(langKey) && configuration.translConfEntryDict[langKey].use)
{
srcToTrgTranslitConf = configuration.translConfEntryDict[langKey];
Log.Write ("Transliteration loaded for language "+srcLang+" into language "+trgLang+": "+ configuration.translConfEntryDict[langKey].mosesIniPath, LogLevelType.LIMITED_OUTPUT,configuration);
foundAtLeastOne = true;
}
else //If the direct dictionary does not exist, log a warning and continue.
{
Log.Write ("Direct transliteration for the pair "+langKey+" was not found.",LogLevelType.WARNING,configuration);
}
//If EN interlingua should not be used or one of the interlingua transliteration configurations is missing, try the direct transliteration also in a reverse direction.
if (configuration.translConfEntryDict.ContainsKey(langKey2) && configuration.translConfEntryDict[langKey2].use)
{
trgToSrcTranslitConf = configuration.translConfEntryDict[langKey2];
Log.Write ("Transliteration loaded for language "+trgLang+" into language "+srcLang+": "+ configuration.translConfEntryDict[langKey2].mosesIniPath, LogLevelType.LIMITED_OUTPUT,configuration);
foundAtLeastOne = true;
}
else //If the direct dictionary does not exist, log a warning and continue.
{
Log.Write ("Direct transliteration for the pair "+langKey2+" was not found.",LogLevelType.WARNING,configuration);
}
if (!foundAtLeastOne)
{
Log.Write ("Direct transliteration for the pairs "+langKey+" nor "+langKey2+" were not found. Will try falling back to interlingua transliteration.",LogLevelType.WARNING,configuration);
bool interlinguaTranslitLoaded = false;
if (configuration.translConfEntryDict.ContainsKey(srcLangKey) && configuration.translConfEntryDict[srcLangKey].use)
{
srcTranslitConf= configuration.translConfEntryDict[srcLangKey];
interlinguaTranslitLoaded = true;
Log.Write ("EN interlingua transliteration loaded for language "+srcLang+": "+ configuration.translConfEntryDict[srcLangKey].mosesIniPath,LogLevelType.LIMITED_OUTPUT,configuration);
}
if (configuration.translConfEntryDict.ContainsKey(trgLangKey) && configuration.translConfEntryDict[trgLangKey].use)
{
trgTranslitConf= configuration.translConfEntryDict[trgLangKey];
interlinguaTranslitLoaded = true;
Log.Write ("EN interlingua transliteration loaded for language "+trgLang+": "+ configuration.translConfEntryDict[trgLangKey].mosesIniPath,LogLevelType.LIMITED_OUTPUT,configuration);
}
if (interlinguaTranslitLoaded)
{
return true;
}else if (configuration.forceEnTranslitInterlingua){
Log.Write ("Cannot force EN interlingua transliteration for the pair "+langKey+" as at least one of the interlingua transliteration configurations is disabled or missing!",LogLevelType.WARNING,configuration);
Log.Write ("The system will be executed without transliteration.",LogLevelType.WARNING,configuration);
}
}
return false;
}
public static void Main(string[] args)
{
string configFile = null;
string method = null;
string inputFile = null;
string inputFormat = "tagged_plaintext";//Allowed values: tagged_plaintext, preprocessed_terms, term_list
string srcInputFile = null;
string trgInputFile = null;
string srcLang = null;
string trgLang = null;
string outputFile = null;
string consolidatedOutputFile = null;
string outputFormat = "";//"tabsep";//Allowed values: ref_tabsep, tabsep, xml
string preProcessedTermOutputFile = null;//"/home/marcis/Dropbox/MonoProjects/MPAligner/MPAligner/bin/Debug/testTermData.xml";//null;
string tempTranslitFile = null;
bool consolidateResults = false;
double consolidationThreshold = 0;
//bool logPrepData = false;
string domainId = "";
string collectionId = "";
//The skipping parameters are just for debugging. Use them only manually!
string skipSrc = "";
string skipTrg = "";
MPAlignerConfiguration configuration = null;
//Read all configuration parameters from the command line.
for (int i=0; i<args.Length; i++) {
if ((args [i] == "-c" || args [i] == "--configuration") && args.Length > i + 1) {
configFile = args [i + 1];
configuration = new MPAlignerConfiguration ();
configuration.Load (configFile);
} else if ((args [i] == "-m" || args [i] == "--method") && args.Length > i + 1) {
method = args [i + 1];
} else if ((args [i] == "-i" || args [i] == "--input-file") && args.Length > i + 1) {
inputFile = args [i + 1];
//} else if (args [i] == "-lp" || args [i] == "--log-pre-processed") {
// logPrepData = true;
} else if ((args [i] == "-if" || args [i] == "--input-format") && args.Length > i + 1) {
inputFormat = args [i + 1];
} else if ((args [i] == "-si" || args [i] == "--source-input") && args.Length > i + 1) {
srcInputFile = args [i + 1];
} else if ((args [i] == "-ti" || args [i] == "--target-input") && args.Length > i + 1) {
trgInputFile = args [i + 1];
} else if ((args [i] == "-sl" || args [i] == "--source-language") && args.Length > i + 1) {
srcLang = MPFramework.MPFrameworkFunctions.GetValidLangString (args [i + 1]);
} else if ((args [i] == "-tl" || args [i] == "--target-language") && args.Length > i + 1) {
trgLang = MPFramework.MPFrameworkFunctions.GetValidLangString (args [i + 1]);
} else if ((args [i] == "-o" || args [i] == "--output-file") && args.Length > i + 1) {
outputFile = args [i + 1];
} else if ((args [i] == "-of" || args [i] == "--output-format") && args.Length > i + 1) {
outputFormat = args [i + 1];
} else if ((args [i] == "-pto" || args [i] == "--pre-processed-term-output-file") && args.Length > i + 1) {
preProcessedTermOutputFile = args [i + 1];
} else if ((args [i] == "-ttf" || args [i] == "--temp-translit-file") && args.Length > i + 1) {
tempTranslitFile = args [i + 1];
} else if ((args [i] == "-ss" || args [i] == "--skip-source-file") && args.Length > i + 1) {
skipSrc = args [i + 1];
} else if ((args [i] == "-st" || args [i] == "--skip-target-file") && args.Length > i + 1) {
skipTrg = args [i + 1];
} else if ((args [i] == "-d_id" || args [i] == "--domain-id") && args.Length > i + 1) {
domainId = args [i + 1];
} else if ((args [i] == "-c_id" || args [i] == "--collection-id") && args.Length > i + 1) {
collectionId = args [i + 1];
} else if ((args [i] == "-ct" || args [i] == "--consolidation-threshold") && args.Length > i + 1) {
//Consolidation works only if the ref_tabsep output format is specified!
NumberFormatInfo nfi = new NumberFormatInfo ();
nfi.CurrencyDecimalSeparator = ".";
nfi.NumberDecimalSeparator = ".";
nfi.PercentDecimalSeparator = ".";
consolidationThreshold = Convert.ToDouble (args [i + 1], nfi);
consolidateResults = true;
}
}
//Break if a method is not defined.
if (string.IsNullOrWhiteSpace (method)) {
Log.Write ("Method not specified!",LogLevelType.ERROR,configuration);
PrintUsage ();
return;
}
//Write a configuration file to the output file if the config method is specified.
if (method.ToLower () == "config") {
if (string.IsNullOrWhiteSpace (outputFile)) {
Log.Write("Output file not specified!",LogLevelType.ERROR,configuration);
PrintUsage ();
return;
}
MPAlignerConfiguration conf = new MPAlignerConfiguration ();
MPAlignerConfigurationDictEntry cde = new MPAlignerConfigurationDictEntry ();
cde.srcLang = "lv";
cde.trgLang = "en";
cde.path = "/home/marcis/TILDE/RESOURCES/DICT/lv_en_noisy";
conf.dictConfEntryDict.Add ("lv_en", cde);
cde = new MPAlignerConfigurationDictEntry ();
cde.srcLang = "lt";
cde.trgLang = "en";
cde.path = "/home/marcis/TILDE/RESOURCES/DICT/lt_en";
conf.dictConfEntryDict.Add ("lt_en", cde);
MPAlignerConfigurationTranslEntry cte = new MPAlignerConfigurationTranslEntry ();
cte.mosesIniPath = "/home/marcis/TILDE/RESOURCES/TRANSLIT_WORKING_DIR/LV-EN/lv-en-binarised-model.moses.ini";
cte.srcLang = "lv";
cte.trgLang = "en";
conf.translConfEntryDict.Add ("lv_en", cte);
cte = new MPAlignerConfigurationTranslEntry ();
cte.mosesIniPath = "/home/marcis/TILDE/RESOURCES/TRANSLIT_WORKING_DIR/LV-EN/lt-en-binarised-model.moses.ini";
cte.srcLang = "lt";
cte.trgLang = "en";
conf.translConfEntryDict.Add ("lt_en", cte);
MPAlignerConfigurationLangPairEntry lpe = new MPAlignerConfigurationLangPairEntry ();
lpe.srcLang = "lv";
lpe.trgLang = "en";
conf.langPairEntryDict.Add ("lv_en", lpe);
lpe = new MPAlignerConfigurationLangPairEntry ();
lpe.srcLang = "lt";
lpe.trgLang = "en";
conf.langPairEntryDict.Add ("lt_en", lpe);
MPAlignerConfigurationExceptionEntry cee = new MPAlignerConfigurationExceptionEntry ();
cee.srcLang = "lv";
cee.trgLang = "en";
cee.path = "/home/marcis/TILDE/RESOURCES/EXC_DICT/lv_en_exc";
conf.excDictEntryDict.Add ("lv_en", cee);
cee = new MPAlignerConfigurationExceptionEntry ();
cee.srcLang = "lt";
cee.trgLang = "en";
cee.path = "/home/marcis/TILDE/RESOURCES/EXC_DICT/lt_en_exc";
conf.excDictEntryDict.Add ("lt_en", cee);
MPAlignerConfigurationStopWordListEntry cswle = new MPAlignerConfigurationStopWordListEntry ();
cswle.lang = "lv";
cswle.path = "/home/marcis/TILDE/RESOURCES/STOP_WORD/lv_stop";
conf.stopWordListEntryDict.Add ("lv", cswle);
cswle = new MPAlignerConfigurationStopWordListEntry ();
cswle.lang = "lt";
cswle.path = "/home/marcis/TILDE/RESOURCES/STOP_WORD/lt_stop";
conf.stopWordListEntryDict.Add ("lt", cswle);
cswle = new MPAlignerConfigurationStopWordListEntry ();
cswle.lang = "en";
cswle.path = "/home/marcis/TILDE/RESOURCES/STOP_WORD/en_stop";
conf.stopWordListEntryDict.Add ("en", cswle);
conf.Save (outputFile);
return;
}
//Try reading the default configuration if none is passed, but if the default configuration can not be found, break.
if (string.IsNullOrWhiteSpace (configFile) && File.Exists ("MPAlignerConfig.xml")) {
configuration = new MPAlignerConfiguration ();
configuration.Load (configFile);
} else if (string.IsNullOrWhiteSpace (configFile)) {
Log.Write("Configuration file missing in application directory and a substitution runtime configuration file is not specified!",LogLevelType.ERROR,configuration);
PrintUsage ();
return;
}
//In the case if an output format is not defined in the command line, read it from the configuration file.
if (string.IsNullOrWhiteSpace (outputFormat))
outputFormat = configuration.outputFormat;
//In the case if the configuration does not specify an output format, use the default output format.
if (string.IsNullOrWhiteSpace (outputFormat)) {
outputFormat = "ref_tabsep";
}
Log.confLogLevel = configuration.logLevel;
if (string.IsNullOrWhiteSpace (tempTranslitFile)) {
tempTranslitFile = outputFile+".tmp";
}
Log.Write ("configFile: "+(configFile!=null?configFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("method: "+(method!=null?method:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("inputFile: "+(inputFile!=null?inputFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("inputFormat: "+(inputFormat!=null?inputFormat:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("srcInputFile: "+(srcInputFile!=null?srcInputFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("trgInputFile: "+(trgInputFile!=null?trgInputFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("srcLang: "+(srcLang!=null?srcLang:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("trgLang: "+(trgLang!=null?trgLang:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("outputFile: "+(outputFile!=null?outputFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("outputFormat: "+(outputFormat!=null?outputFormat:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("preProcessedTermOutputFile: "+(preProcessedTermOutputFile!=null?preProcessedTermOutputFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("tempTranslitFile: "+(tempTranslitFile!=null?tempTranslitFile:""),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("consolidation threshold: "+(consolidateResults?consolidationThreshold.ToString():""),LogLevelType.LIMITED_OUTPUT,configuration);
if (outputFormat == "ref_tabsep" && consolidateResults) {
consolidatedOutputFile = outputFile;
outputFile += ".raw";
}
//For document pair-based alignment.
if (method.ToLower () == "taggedfilepairs") {
char[] sep = {'\t'};
if (string.IsNullOrWhiteSpace(inputFile)||!File.Exists(inputFile))
{
Log.Write("Input file list file not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source and/or target languages not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
//Read the alignment thresholds and other language pair specific numerical/single-value parameters.
MPAlignerConfigurationLangPairEntry lpeConf = ReadLangPairConfig (srcLang, trgLang, configuration);
//The size of the cache may affect the performance of the alignment!
Dictionary<string, ProcessedTermEntry> srcTermCache = new Dictionary<string, ProcessedTermEntry>();
Dictionary<string, ProcessedTermEntry> trgTermCache = new Dictionary<string, ProcessedTermEntry>();
bool interlinguaDictUsed = false;
bool interlinguaTranslitUsed = false;
//Define dictionaries for pre-processing.
Dictionary<string, Dictionary<string, double>> srcDict = null;
Dictionary<string, Dictionary<string, double>> trgDict = null;
Dictionary<string, Dictionary<string, double>> srcToTrgDict = null;
Dictionary<string, Dictionary<string, double>> trgToSrcDict = null;
//Define transliteration configurations for pre-processing.
MPAlignerConfigurationTranslEntry srcTranslitConf = null;
MPAlignerConfigurationTranslEntry trgTranslitConf = null;
MPAlignerConfigurationTranslEntry srcToTrgTranslitConf = null;
MPAlignerConfigurationTranslEntry trgToSrcTranslitConf = null;
//Read dictionaries and transliterations.
interlinguaDictUsed = ReadDictionaries(configuration,srcLang,trgLang, out srcDict, out trgDict, out srcToTrgDict, out trgToSrcDict);
interlinguaTranslitUsed = GetTranslitConfig(configuration,srcLang,trgLang,out srcTranslitConf,out trgTranslitConf,out srcToTrgTranslitConf, out trgToSrcTranslitConf);
//Define the alignments (the variable holding alignment results)
Dictionary<string,Dictionary<string, AlignmentInfoElement>> alignments = new Dictionary<string, Dictionary<string, AlignmentInfoElement>>();
//Define and read exception dictionaries.
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
//Define and read stopword lists.
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
StreamReader sr = new StreamReader(inputFile,Encoding.UTF8);
int pairCounter = 0;
bool skip = !string.IsNullOrWhiteSpace(skipSrc)&&!string.IsNullOrWhiteSpace(skipTrg)?true:false;
//Read input document alignment file and process file pairs.
while(!sr.EndOfStream)
{
pairCounter++;
string line = sr.ReadLine().Trim();
if (string.IsNullOrWhiteSpace(line)) continue;
string[] arr = line.Split(sep, StringSplitOptions.RemoveEmptyEntries);
if (arr.Length<2)
{
continue; //If the alignment line does not contain at least two entries, the document alignment is not valid.
}
string srcFile = arr[0];
string trgFile = arr[1];
if (!File.Exists(srcFile))
{
Log.Write("Input file \""+srcFile+"\" cannot be found!",LogLevelType.WARNING,configuration);
continue;
}
if (!File.Exists(trgFile))
{
Log.Write("Input file \""+trgFile+"\" cannot be found!",LogLevelType.WARNING,configuration);
continue;
}
string srcFileName = Path.GetFileName(srcFile);
string trgFileName = Path.GetFileName(trgFile);
//The skipping condition is for debugging - if the system crashes due to insufficient memory...
if (skip)
{
if (srcFileName==skipSrc&&trgFileName == skipTrg)
{
skip = false;
}
else
{
Log.Write("Skipping file pair "+srcFileName+" and " + trgFileName+".",LogLevelType.WARNING,configuration);
continue;
}
}
Log.Write("Processing file pair "+srcFileName+" and " + trgFileName+".",LogLevelType.LIMITED_OUTPUT,configuration);
//Define term entry data variables (used for sotring terms in pre-pre-processed and pre-processed states).
Dictionary<string,SimpleTermEntry> srcInitialList = new Dictionary<string, SimpleTermEntry>();
Dictionary<string,SimpleTermEntry> trgInitialList = new Dictionary<string, SimpleTermEntry>();
Dictionary<string,SimpleTermEntry> srcInitialTempList = new Dictionary<string, SimpleTermEntry>();
Dictionary<string,SimpleTermEntry> trgInitialTempList = new Dictionary<string, SimpleTermEntry>();
Dictionary<string, ProcessedTermEntry> srcTermList = new Dictionary<string, ProcessedTermEntry>();
Dictionary<string, ProcessedTermEntry> trgTermList = new Dictionary<string, ProcessedTermEntry>();
Dictionary<string, ProcessedTermEntry> srcTermTempList = new Dictionary<string, ProcessedTermEntry>();
Dictionary<string, ProcessedTermEntry> trgTermTempList = new Dictionary<string, ProcessedTermEntry>();
//Two input formats are currently supported - term-tagged plaintext files and term list (one term per line) files.
if (inputFormat=="tagged_plaintext")
{
//Read terms from the term-tagged documents.
srcInitialTempList = TermTaggedFileParser.ParseTermTaggedFile(srcFile,Encoding.UTF8, configuration.concLen);
trgInitialTempList = TermTaggedFileParser.ParseTermTaggedFile(trgFile,Encoding.UTF8, configuration.concLen);
}
else
{
//Read terms from the term list files.
srcInitialTempList = ListFileParser.Parse(srcFile,Encoding.UTF8);
trgInitialTempList = ListFileParser.Parse(trgFile,Encoding.UTF8);
}
//Search for already pre-processed source terms in the cache.
foreach(string term in srcInitialTempList.Keys)
{
string lower = term.ToLower();
if (srcTermCache.ContainsKey(lower))
{
if (!srcTermList.ContainsKey(lower)) srcTermList.Add(lower, srcTermCache[lower]);
}
else
{
srcInitialList.Add(term, srcInitialTempList[term]);
}
}
//Search for already pre-processed target terms in the cache.
foreach(string term in trgInitialTempList.Keys)
{
string lower = term.ToLower();
if (trgTermCache.ContainsKey(lower))
{
if (!trgTermList.ContainsKey(lower)) trgTermList.Add(lower, trgTermCache[lower]);
}
else
{
trgInitialList.Add(term, trgInitialTempList[term]);
}
}
//Now pre-process terms that have not been pre-processed again.
if (srcDict!=null||trgDict!=null)
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermTempList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermTempList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermTempList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermTempList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
else
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermTempList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermTempList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermTempList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermTempList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
//Update the pre-processed term list for alignment.
foreach(string srcTerm in srcTermTempList.Keys)
{
if (!srcTermList.ContainsKey(srcTerm)) srcTermList.Add(srcTerm,srcTermTempList[srcTerm]);
if (!srcTermCache.ContainsKey(srcTerm)) srcTermCache.Add(srcTerm, srcTermTempList[srcTerm]);
}
foreach(string trgTerm in trgTermTempList.Keys)
{
if (!trgTermList.ContainsKey(trgTerm)) trgTermList.Add(trgTerm,trgTermTempList[trgTerm]);
if (!trgTermCache.ContainsKey(trgTerm)) trgTermCache.Add(trgTerm, trgTermTempList[trgTerm]);
}
//Execute alignment for one file pair.
List<AlignmentInfoElement> alignment = new List<AlignmentInfoElement>();
//The execution may be multi-threaded or single-threaded. The multi-threaded execution may be instable. Therefore, be careful when using multi-threading.
if (configuration.useMultiThreadedExecution)
{
alignment = Alignment.AlignPairsMultiThreaded(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcFile, trgFile, excDict, srcStopWords, trgStopWords);
}
else
{
alignment = Alignment.AlignPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcFile, trgFile, excDict, srcStopWords, trgStopWords);
}
if (alignment!=null)
{
foreach(AlignmentInfoElement aie in alignment)
{
if (!alignments.ContainsKey(aie.srcEntry.lowercaceForm))
{
alignments.Add(aie.srcEntry.lowercaceForm, new Dictionary<string, AlignmentInfoElement>());
}
if (!alignments[aie.srcEntry.lowercaceForm].ContainsKey(aie.trgEntry.lowercaceForm))
{
alignments[aie.srcEntry.lowercaceForm].Add(aie.trgEntry.lowercaceForm, aie);
}
}
}
//If pre-processed term cache is full, empty it (this maybe can be imrpoved with the help of some sort of a flowing cache (always circulating).
if (srcTermCache.Count>50000)
{
srcTermCache.Clear();
srcTermCache = new Dictionary<string, ProcessedTermEntry>();
GC.Collect();
GC.WaitForPendingFinalizers();
}
if (trgTermCache.Count>50000)
{
trgTermCache.Clear();
trgTermCache = new Dictionary<string, ProcessedTermEntry>();
GC.Collect();
GC.WaitForPendingFinalizers();
}
//After each 50 pairs, print rsults.
if (pairCounter%50==0||alignments.Count>50000)
{
Log.Write("Printing intermediate results after "+pairCounter.ToString()+" file pairs",LogLevelType.LIMITED_OUTPUT,configuration);
List<AlignmentInfoElement> resAlignment = new List<AlignmentInfoElement>();
foreach(string src in alignments.Keys)
{
foreach(string trg in alignments[src].Keys)
{
resAlignment.Add(alignments[src][trg]);
}
}
AlignmentInfoElement.AppendList(outputFormat,outputFile,resAlignment,lpeConf,srcLang,trgLang,collectionId,domainId);
alignments.Clear();
alignments = new Dictionary<string, Dictionary<string, AlignmentInfoElement>>();
GC.Collect();
GC.WaitForPendingFinalizers();
}
}
sr.Close();
//If there are alignments left, write them to the output file.
if (!string.IsNullOrWhiteSpace(outputFile))
{
Log.Write("Printing final results after "+pairCounter.ToString()+" file pairs",LogLevelType.LIMITED_OUTPUT,configuration);
List<AlignmentInfoElement> resAlignment = new List<AlignmentInfoElement>();
foreach(string src in alignments.Keys)
{
foreach(string trg in alignments[src].Keys)
{
resAlignment.Add(alignments[src][trg]);
}
}
AlignmentInfoElement.AppendList(outputFormat,outputFile,resAlignment,lpeConf,srcLang,trgLang,collectionId,domainId);
}
}
else if (method.ToLower () == "singletaggedpair") //TODO: REFACTOR (the file pair list processing could be handled (wisely) through a single file pair processing method!!!
{
//Define the instances of source and target processed term lists.
Dictionary<string, ProcessedTermEntry> srcTermList = new Dictionary<string, ProcessedTermEntry>();
Dictionary<string, ProcessedTermEntry> trgTermList = new Dictionary<string, ProcessedTermEntry>();
bool interlinguaDictUsed = false;
bool interlinguaTranslitUsed = false;
if (inputFormat=="preprocessed_terms")
{
if (string.IsNullOrWhiteSpace(inputFile)||!File.Exists(inputFile))
{
Log.Write("Pre-processed term input file not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source and/or target languages not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
PreprocessedTermData ptd = PreprocessedTermData.ReadFromFile(inputFile);
foreach(ProcessedTermEntry pte in ptd.srcTerms)
{
if(!srcTermList.ContainsKey(pte.lowercaceForm))
{
srcTermList.Add(pte.lowercaceForm,pte);
}
}
foreach(ProcessedTermEntry pte in ptd.trgTerms)
{
if(!trgTermList.ContainsKey(pte.lowercaceForm))
{
trgTermList.Add(pte.lowercaceForm,pte);
}
}
srcLang = ptd.srcLang;
trgLang = ptd.trgLang;
interlinguaDictUsed = ptd.interlinguaDictUsed;
interlinguaTranslitUsed = ptd.interlinguaTranslitUsed;
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
if (!string.IsNullOrWhiteSpace(outputFile))
{
List<AlignmentInfoElement> alignment = new List<AlignmentInfoElement>();
if (configuration.useMultiThreadedExecution)
{
alignment = Alignment.AlignPairsMultiThreaded(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
else
{
alignment = Alignment.AlignPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
AlignmentInfoElement.PrintList(outputFormat,outputFile,alignment, configuration.printTopTrgForSrc,null,srcLang,trgLang,collectionId,domainId);
}
}
else if (inputFormat=="term_list"||inputFormat=="tagged_plaintext")
{
if (string.IsNullOrWhiteSpace(srcInputFile)||!File.Exists(srcInputFile)||string.IsNullOrWhiteSpace(trgInputFile)||!File.Exists(trgInputFile))
{
Log.Write("Source and/or target files not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source and/or target languages not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
Dictionary<string,SimpleTermEntry> srcInitialList = new Dictionary<string, SimpleTermEntry>();
Dictionary<string,SimpleTermEntry> trgInitialList = new Dictionary<string, SimpleTermEntry>();
if (inputFormat=="tagged_plaintext")
{
srcInitialList = TermTaggedFileParser.ParseTermTaggedFile(srcInputFile,Encoding.UTF8, configuration.concLen);
trgInitialList = TermTaggedFileParser.ParseTermTaggedFile(trgInputFile,Encoding.UTF8, configuration.concLen);
}
else
{
srcInitialList = ListFileParser.Parse(srcInputFile,Encoding.UTF8);
trgInitialList = ListFileParser.Parse(trgInputFile,Encoding.UTF8);
}
Log.Write ("Unprocessed source terms: "+srcInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Unprocessed target terms: "+trgInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Dictionary<string, Dictionary<string, double>> srcDict = null;
Dictionary<string, Dictionary<string, double>> trgDict = null;
Dictionary<string, Dictionary<string, double>> srcToTrgDict = null;
Dictionary<string, Dictionary<string, double>> trgToSrcDict = null;
MPAlignerConfigurationTranslEntry srcTranslitConf = null;
MPAlignerConfigurationTranslEntry trgTranslitConf = null;
MPAlignerConfigurationTranslEntry srcToTrgTranslitConf = null;
MPAlignerConfigurationTranslEntry trgToSrcTranslitConf = null;
interlinguaDictUsed = ReadDictionaries(configuration,srcLang,trgLang, out srcDict, out trgDict, out srcToTrgDict, out trgToSrcDict);
interlinguaTranslitUsed = GetTranslitConfig(configuration,srcLang,trgLang,out srcTranslitConf,out trgTranslitConf,out srcToTrgTranslitConf, out trgToSrcTranslitConf);
if (srcDict!=null||trgDict!=null)
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
else
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
Log.Write ("Pre-processed source terms: "+srcTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Pre-processed target terms: "+trgTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
///If pre-processed terms should be saved for future use an output format is created.
/// This functionality is not available for the file pair list-based processing.
if (!string.IsNullOrWhiteSpace(preProcessedTermOutputFile))
{
List<ProcessedTermEntry> srcTerms = new List<ProcessedTermEntry>(srcTermList.Values);
List<ProcessedTermEntry> trgTerms = new List<ProcessedTermEntry>(trgTermList.Values);
PreprocessedTermData ptd = new PreprocessedTermData();
ptd.interlinguaDictUsed = interlinguaDictUsed;
ptd.interlinguaTranslitUsed = interlinguaTranslitUsed;
ptd.srcTerms = srcTerms.ToArray();
ptd.trgTerms = trgTerms.ToArray();
ptd.srcLang = srcLang;
ptd.trgLang = trgLang;
string outStr = MPFramework.MPFrameworkFunctions.SerializeObjectInstance<PreprocessedTermData>(ptd);
File.WriteAllText(preProcessedTermOutputFile,outStr);
}
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
if (!string.IsNullOrWhiteSpace(outputFile))
{
List<AlignmentInfoElement> alignment = new List<AlignmentInfoElement>();
if (configuration.useMultiThreadedExecution)
{
alignment = Alignment.AlignPairsMultiThreaded(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
else
{
alignment = Alignment.AlignPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
AlignmentInfoElement.PrintList(outputFormat,outputFile,alignment, configuration.printTopTrgForSrc,null,srcLang,trgLang,collectionId,domainId);
}
}
else
{
Log.Write ("Input format UNKNOWN or UNDEFINED.",LogLevelType.ERROR,configuration);
return;
}
}
else if (method.ToLower () == "singletermpairlist") //Use this method only if filtering of term pairs or some sort of evaluation is necessary!
{
//Define the instances of source and target processed term lists.
List<ProcessedTermEntry> srcTermList = new List<ProcessedTermEntry>();
List<ProcessedTermEntry> trgTermList = new List<ProcessedTermEntry>();
bool interlinguaDictUsed = false;
bool interlinguaTranslitUsed = false;
if (inputFormat=="preprocessed_terms")
{
if (string.IsNullOrWhiteSpace(inputFile)||!File.Exists(inputFile))
{
Log.Write("Pre-processed term input file not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source and/or target languages not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
PreprocessedTermData ptd = PreprocessedTermData.ReadFromFile(inputFile);
srcTermList.AddRange(ptd.srcTerms);
trgTermList.AddRange(ptd.trgTerms);
Log.Write ("Pre-processed source terms: "+srcTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Pre-processed target terms: "+trgTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
srcLang = ptd.srcLang;
trgLang = ptd.trgLang;
interlinguaDictUsed = ptd.interlinguaDictUsed;
interlinguaTranslitUsed = ptd.interlinguaTranslitUsed;
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
Log.Write ("Exception dictionary entries: "+excDict.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Log.Write ("Source language stopwords: "+srcStopWords.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
Log.Write ("Target language stopwords: "+trgStopWords.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
if (!string.IsNullOrWhiteSpace(outputFile))
{
List<AlignmentInfoElement> alignment = Alignment.AlignListPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
Log.Write ("Alignment elements after alignment: "+alignment.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
AlignmentInfoElement.PrintList(outputFormat,outputFile,alignment, configuration.printTopTrgForSrc,null,srcLang,trgLang,collectionId,domainId);
}
}
else
{
if (string.IsNullOrWhiteSpace(srcInputFile)||!File.Exists(srcInputFile)||string.IsNullOrWhiteSpace(trgInputFile)||!File.Exists(trgInputFile))
{
Log.Write("Source and/or target files not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source and/or target languages not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
List<string> srcInitialList = new List<string>();
List<string> trgInitialList = new List<string>();
srcInitialList = ListFileParser.ParseList(srcInputFile,Encoding.UTF8);
trgInitialList = ListFileParser.ParseList(trgInputFile,Encoding.UTF8);
if (srcInitialList.Count!=trgInitialList.Count)
{
Log.Write("Source and target term lists are with different lengths",LogLevelType.ERROR,configuration);
throw new ArgumentException("Source and target term lists are with different lengths");
}
Log.Write ("Unprocessed source terms: "+srcInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Unprocessed target terms: "+trgInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Dictionary<string, Dictionary<string, double>> srcDict = null;
Dictionary<string, Dictionary<string, double>> trgDict = null;
Dictionary<string, Dictionary<string, double>> srcToTrgDict = null;
Dictionary<string, Dictionary<string, double>> trgToSrcDict = null;
MPAlignerConfigurationTranslEntry srcTranslitConf = null;
MPAlignerConfigurationTranslEntry trgTranslitConf = null;
MPAlignerConfigurationTranslEntry srcToTrgTranslitConf = null;
MPAlignerConfigurationTranslEntry trgToSrcTranslitConf = null;
interlinguaDictUsed = ReadDictionaries(configuration,srcLang,trgLang, out srcDict, out trgDict, out srcToTrgDict, out trgToSrcDict);
interlinguaTranslitUsed = GetTranslitConfig(configuration,srcLang,trgLang,out srcTranslitConf,out trgTranslitConf,out srcToTrgTranslitConf, out trgToSrcTranslitConf);
if (srcDict!=null||trgDict!=null)
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermList = ProcessedTermEntry.ProcessTermsList(srcInitialList,srcDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTermsList(trgInitialList,trgDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermList = ProcessedTermEntry.ProcessTermsList(srcInitialList,srcDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTermsList(trgInitialList,trgDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
else
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
srcTermList = ProcessedTermEntry.ProcessTermsList(srcInitialList,srcToTrgDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTermsList(trgInitialList,trgToSrcDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
srcTermList = ProcessedTermEntry.ProcessTermsList(srcInitialList,srcToTrgDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
trgTermList = ProcessedTermEntry.ProcessTermsList(trgInitialList,trgToSrcDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
Log.Write ("Pre-processed source terms: "+srcTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Pre-processed target terms: "+trgTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
///If pre-processed terms should be saved for future use an output format is created.
if (!string.IsNullOrWhiteSpace(preProcessedTermOutputFile))
{
PreprocessedTermData ptd = new PreprocessedTermData();
ptd.interlinguaDictUsed = interlinguaDictUsed;
ptd.interlinguaTranslitUsed = interlinguaTranslitUsed;
ptd.srcTerms = srcTermList.ToArray();
ptd.trgTerms = trgTermList.ToArray();
ptd.srcLang = srcLang;
ptd.trgLang = trgLang;
string outStr = MPFramework.MPFrameworkFunctions.SerializeObjectInstance<PreprocessedTermData>(ptd);
File.WriteAllText(preProcessedTermOutputFile,outStr);
}
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
if (!string.IsNullOrWhiteSpace(outputFile))
{
List<AlignmentInfoElement> alignment = Alignment.AlignListPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
AlignmentInfoElement.PrintList(outputFormat,outputFile,alignment, configuration.printTopTrgForSrc,null,srcLang,trgLang,collectionId,domainId);
}
}
}
else if (method.ToLower () == "eurovoceval")
{
if (string.IsNullOrWhiteSpace(inputFile)||!File.Exists(inputFile))
{
Log.Write("Eurovoc input file not specified or cannot be found!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
if (string.IsNullOrWhiteSpace(srcLang)||string.IsNullOrWhiteSpace(trgLang))
{
Log.Write("Source or target language not specified!",LogLevelType.ERROR,configuration);
PrintUsage();
return;
}
configuration.allowTrimmedAlignments = false;
//configuration.useMultiThreadedExecution = false;
configuration.printTopTrgForSrc = true;
string logFile = outputFile+".res.log";
StreamWriter sw = new StreamWriter(logFile, true, Encoding.UTF8);
Dictionary<string,List<string>> eurovocDict = ReadEurovocDict(inputFile);;
//List<string> langList = GetLangsFromConf(configuration);
//for(int i = 0;i<langList.Count;i++)
//{
//for(int j = 0;j<langList.Count;j++)
//{
//if (i==j) continue;
//srcLang = langList[i];
//trgLang = langList[j];
Log.Write("Processing pair "+srcLang+"_"+trgLang,LogLevelType.LIMITED_OUTPUT,configuration);
if (Char.IsDigit(outputFile[outputFile.Length-1])) outputFile = outputFile.Substring(0,outputFile.Length-1);
string alignmentOutputFile = outputFile+"."+srcLang+"_"+trgLang+".align.txt";
if (File.Exists(alignmentOutputFile))
{
Log.Write("Pair "+srcLang+"_"+trgLang+" already processed! Evaluating...",LogLevelType.LIMITED_OUTPUT,configuration);
List<StringComparisonElement> terms = new List<StringComparisonElement>();
StreamReader sr = new StreamReader(alignmentOutputFile,Encoding.UTF8);
char[] sep = {'\t'};
NumberFormatInfo nfi = new NumberFormatInfo();
nfi.CurrencyDecimalSeparator=".";
nfi.NumberDecimalSeparator=".";
nfi.PercentDecimalSeparator=".";
while(!sr.EndOfStream)
{
string line = sr.ReadLine().Trim();
string[] arr = line.Split(sep,StringSplitOptions.None);
if (arr.Length>=3)
{
StringComparisonElement sce = new StringComparisonElement();
sce.src = arr[0];
sce.trg = arr[1];
sce.similarity = Convert.ToDouble(arr[2],nfi);
terms.Add(sce);
}
}
sr.Close();
terms.Sort();
List<double> scores = new List<double>();
double tmp = 0;
while (tmp<=1)
{
scores.Add(tmp);
tmp+=0.01;
}
List<double> correct = new List<double>();
for(int t=0;t<scores.Count;t++)
{
correct.Add(0);
}
List<double> total = new List<double>();
for(int t=0;t<scores.Count;t++)
{
total.Add(0);
}
int totalForRec = 0;
Dictionary<string,Dictionary<string,bool>> goldList = new Dictionary<string, Dictionary<string, bool>>();
for (int s = 0;s<eurovocDict[srcLang].Count;s++)
{
if (!eurovocDict[srcLang][s].Contains("(under translation)")&&!eurovocDict[trgLang][s].Contains("(under translation)"))
{
totalForRec++;
if (!goldList.ContainsKey(eurovocDict[srcLang][s].ToLower())) goldList.Add(eurovocDict[srcLang][s].ToLower(), new Dictionary<string,bool>());
if (!goldList[eurovocDict[srcLang][s].ToLower()].ContainsKey(eurovocDict[trgLang][s].ToLower())) goldList[eurovocDict[srcLang][s].ToLower()].Add(eurovocDict[trgLang][s].ToLower(),true);
}
}
string previousSrc = null;
foreach(StringComparisonElement sce in terms)
{
string currSrc = sce.src;
if (previousSrc!=currSrc.ToLower())
{
string src = sce.src.ToLower();
string trg = sce.trg.ToLower();
double alignScore = sce.similarity;
bool corr = false;
if (goldList.ContainsKey(src)&&goldList[src].ContainsKey(trg)) corr = true;
for (int s =0;s<scores.Count;s++)
{
if (scores[s]<=alignScore)
{
if (corr) correct[s]++;
total[s]++;
}
}
previousSrc = currSrc.ToLower();
}
}
for(int s=0;s<scores.Count;s++)
{
double corr = correct[s];
double tot = total[s];
double totCorr = totalForRec;
double prec = corr/tot*100;
double rec = corr/totCorr*100;
double f1 = prec*rec*2/(prec+rec);
sw.WriteLine(srcLang+"\t"+trgLang+"\t"+scores[s].ToString()+"\t"+corr.ToString()+"\t"+tot.ToString()+"\t"+totCorr.ToString()+"\t"+prec.ToString()+"\t"+rec.ToString()+"\t"+f1.ToString());
}
sw.Flush();
//}
//}
sw.Close();
//continue;
return;
}
string preprocessedOutputFile = outputFile+"."+srcLang+"_"+trgLang+".prep.txt";
Dictionary<string,SimpleTermEntry> srcInitialList = StringListToDict(eurovocDict[srcLang]);
Dictionary<string,SimpleTermEntry> trgInitialList = StringListToDict(eurovocDict[trgLang]);
Log.Write ("Unprocessed source terms: "+srcInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Unprocessed target terms: "+trgInitialList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Dictionary<string, Dictionary<string, double>> srcDict = null;
Dictionary<string, Dictionary<string, double>> trgDict = null;
Dictionary<string, Dictionary<string, double>> srcToTrgDict = null;
Dictionary<string, Dictionary<string, double>> trgToSrcDict = null;
MPAlignerConfigurationTranslEntry srcTranslitConf = null;
MPAlignerConfigurationTranslEntry trgTranslitConf = null;
MPAlignerConfigurationTranslEntry srcToTrgTranslitConf = null;
MPAlignerConfigurationTranslEntry trgToSrcTranslitConf = null;
bool interlinguaDictUsed = ReadDictionaries(configuration,srcLang,trgLang, out srcDict, out trgDict, out srcToTrgDict, out trgToSrcDict);
bool interlinguaTranslitUsed = GetTranslitConfig(configuration,srcLang,trgLang,out srcTranslitConf,out trgTranslitConf,out srcToTrgTranslitConf, out trgToSrcTranslitConf);
Dictionary<string,ProcessedTermEntry> srcTermList = new Dictionary<string,ProcessedTermEntry>();
Dictionary<string,ProcessedTermEntry> trgTermList = new Dictionary<string,ProcessedTermEntry>();
if (File.Exists(preprocessedOutputFile))
{
Log.Write("Preprocessed term data found! Reading pre-processed data to save time!", LogLevelType.WARNING,configuration);
PreprocessedTermData ptd1 = PreprocessedTermData.ReadFromFile(preprocessedOutputFile);
interlinguaDictUsed = ptd1.interlinguaDictUsed;
interlinguaTranslitUsed = ptd1.interlinguaTranslitUsed;
foreach(ProcessedTermEntry pte in ptd1.srcTerms)
{
if (!srcTermList.ContainsKey(pte.lowercaceForm))
{
srcTermList.Add(pte.lowercaceForm,pte);
}
}
foreach(ProcessedTermEntry pte in ptd1.trgTerms)
{
if (!trgTermList.ContainsKey(pte.lowercaceForm))
{
trgTermList.Add(pte.lowercaceForm,pte);
}
}
}
else if (interlinguaDictUsed&&interlinguaTranslitUsed)
{
string dir = Path.GetDirectoryName(preprocessedOutputFile);
if (!dir.EndsWith(Path.DirectorySeparatorChar.ToString())) dir+=Path.DirectorySeparatorChar.ToString();
string prepSrcToTrgFile = dir+"eurovoc_preprocessed_"+srcLang+"_en.xml";
string prepTrgToSrcFile = dir+"eurovoc_preprocessed_"+trgLang+"_en.xml";
if (File.Exists(prepSrcToTrgFile))
{
Log.Write ("Reading processed term list: eurovoc_preprocessed_"+srcLang+"_en.xml",LogLevelType.LIMITED_OUTPUT,configuration);
srcTermList = ProcessedTermEntry.ReadFromFile(prepSrcToTrgFile);
}
if (File.Exists(prepTrgToSrcFile))
{
Log.Write ("Reading processed term list: eurovoc_preprocessed_"+trgLang+"_en.xml",LogLevelType.LIMITED_OUTPUT,configuration);
trgTermList = ProcessedTermEntry.ReadFromFile(prepTrgToSrcFile);
}
}
else if (!interlinguaDictUsed&&!interlinguaTranslitUsed)
{
string dir = Path.GetDirectoryName(preprocessedOutputFile);
if (!dir.EndsWith(Path.DirectorySeparatorChar.ToString())) dir+=Path.DirectorySeparatorChar.ToString();
string prepSrcToTrgFile = dir+"eurovoc_preprocessed_"+srcLang+"_"+trgLang+".xml";
string prepTrgToSrcFile = dir+"eurovoc_preprocessed_"+trgLang+"_"+srcLang+".xml";
if (File.Exists(prepSrcToTrgFile))
{
Log.Write ("Reading processed term list: eurovoc_preprocessed_"+srcLang+"_"+trgLang+".xml",LogLevelType.LIMITED_OUTPUT,configuration);
srcTermList = ProcessedTermEntry.ReadFromFile(prepSrcToTrgFile);
}
if (File.Exists(prepTrgToSrcFile))
{
Log.Write ("Reading processed term list: eurovoc_preprocessed_"+trgLang+"_"+srcLang+".xml",LogLevelType.LIMITED_OUTPUT,configuration);
trgTermList = ProcessedTermEntry.ReadFromFile(prepTrgToSrcFile);
}
}
if (srcDict!=null||trgDict!=null)
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
if (srcTermList.Count<1)
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
if (trgTermList.Count<1)
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
if (srcTermList.Count<1)
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
if (trgTermList.Count<1)
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
else
{
if (srcTranslitConf!=null && trgTranslitConf!=null)
{
if (srcTermList.Count<1)
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
if (trgTermList.Count<1)
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
else
{
if (srcTermList.Count<1)
srcTermList = ProcessedTermEntry.ProcessTerms(srcInitialList,srcToTrgDict,srcLang,srcToTrgTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
if (trgTermList.Count<1)
trgTermList = ProcessedTermEntry.ProcessTerms(trgInitialList,trgToSrcDict,trgLang,trgToSrcTranslitConf, configuration.mosesPath, tempTranslitFile,configuration.alignmentThreads);
}
}
Log.Write ("Pre-processed source terms: "+srcTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
Log.Write ("Pre-processed target terms: "+trgTermList.Count.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
///If pre-processed terms should be saved for future use an output format is created.
List<ProcessedTermEntry> srcTerms = new List<ProcessedTermEntry>(srcTermList.Values);
List<ProcessedTermEntry> trgTerms = new List<ProcessedTermEntry>(trgTermList.Values);
PreprocessedTermData ptd = new PreprocessedTermData();
ptd.interlinguaDictUsed = interlinguaDictUsed;
ptd.interlinguaTranslitUsed = interlinguaTranslitUsed;
ptd.srcTerms = srcTerms.ToArray();
ptd.trgTerms = trgTerms.ToArray();
ptd.srcLang = srcLang;
ptd.trgLang = trgLang;
string outStr = MPFramework.MPFrameworkFunctions.SerializeObjectInstance<PreprocessedTermData>(ptd);
File.WriteAllText(preprocessedOutputFile,outStr);
Dictionary<string, Dictionary<string, bool>> excDict = null;
ReadExceptionDictionary(configuration,srcLang, trgLang,out excDict);
Dictionary<string,bool> srcStopWords = null;
ReadStopwordList(configuration,srcLang,out srcStopWords);
Dictionary<string,bool> trgStopWords = null;
ReadStopwordList(configuration,trgLang,out trgStopWords);
//Need to pre-set the alignment thresholds, otherwise these will be overriden by defaults.
MPAlignerConfigurationLangPairEntry lpeConf = ReadLangPairConfig (srcLang, trgLang, configuration);
List<AlignmentInfoElement> alignment = new List<AlignmentInfoElement>();
if (configuration.useMultiThreadedExecution)
{
alignment = Alignment.AlignPairsMultiThreaded(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
else
{
alignment = Alignment.AlignPairs(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
}
//Multi-threaded execution is not stable at the moment...
//List<AlignmentInfoElement> alignment = Alignment.AlignPairsMultiThreaded(configuration,srcTermList,trgTermList,interlinguaDictUsed,interlinguaTranslitUsed,srcLang,trgLang, srcInputFile, trgInputFile, excDict, srcStopWords, trgStopWords);
AlignmentInfoElement.PrintList(outputFormat, alignmentOutputFile, alignment, configuration.printTopTrgForSrc, lpeConf,srcLang,trgLang,collectionId,domainId);
{
List<double> scores = new List<double>();
double tmp = 0;
while (tmp<=1)
{
scores.Add(tmp);
tmp+=0.01;
}
List<double> correct = new List<double>();
for(int t=0;t<scores.Count;t++)
{
correct.Add(0);
}
List<double> total = new List<double>();
for(int t=0;t<scores.Count;t++)
{
total.Add(0);
}
int totalForRec = 0;
Dictionary<string,Dictionary<string,bool>> goldList = new Dictionary<string, Dictionary<string, bool>>();
for (int s = 0;s<eurovocDict[srcLang].Count;s++)
{
if (!eurovocDict[srcLang][s].ToLower().Contains("(under translation)")&&!eurovocDict[trgLang][s].ToLower().Contains("(under translation)"))
{
totalForRec++;
if (!goldList.ContainsKey(eurovocDict[srcLang][s].ToLower())) goldList.Add(eurovocDict[srcLang][s].ToLower(), new Dictionary<string,bool>());
if (!goldList[eurovocDict[srcLang][s].ToLower()].ContainsKey(eurovocDict[trgLang][s].ToLower())) goldList[eurovocDict[srcLang][s].ToLower()].Add(eurovocDict[trgLang][s].ToLower(),true);
}
}
string previousSrc = null;
alignment.Sort();
foreach(AlignmentInfoElement aie in alignment)
{
string currSrc = AlignmentInfoElement.GetStrFromEntry(aie.srcEntry.surfaceFormWords, aie.minSrcId, aie.maxSrcId);
if (previousSrc!=currSrc.ToLower())
{
string src = aie.srcEntry.surfaceForm.ToLower();
string trg = aie.trgEntry.surfaceForm.ToLower();
double alignScore = aie.alignmentScore;
bool corr = false;
if (goldList.ContainsKey(src)&&goldList[src].ContainsKey(trg)) corr = true;
for (int s =0;s<scores.Count;s++)
{
if (scores[s]<=alignScore)
{
if (corr) correct[s]++;
total[s]++;
}
}
previousSrc = currSrc.ToLower();
}
}
for(int s=0;s<scores.Count;s++)
{
double corr = correct[s];
double tot = total[s];
double totCorr = totalForRec;
double prec = corr/tot*100;
double rec = corr/totCorr*100;
double f1 = prec*rec*2/(prec+rec);
sw.WriteLine(srcLang+"\t"+trgLang+"\t"+scores[s].ToString()+"\t"+corr.ToString()+"\t"+tot.ToString()+"\t"+totCorr.ToString()+"\t"+prec.ToString()+"\t"+rec.ToString()+"\t"+f1.ToString());
}
sw.Flush();
//}
//}
sw.Close();
}
}
if (File.Exists(tempTranslitFile)) File.Delete(tempTranslitFile);
if (consolidateResults) {
Log.Write ("Consolidating aligned term pairs with a threshold of: "+consolidationThreshold.ToString(),LogLevelType.LIMITED_OUTPUT,configuration);
//In the case if -ct (consolidation threshold) was defined and the output format has been ref_tabsep, the consolidation of results is perfomed.
ConsolidationElement.ConsolidateRefTabsep(outputFile, consolidatedOutputFile,consolidationThreshold);
}
}
