public MarianCustomizer(
MTModel model,
MTModel customModel,
ParallelFilePair inputPair,
ParallelFilePair indomainValidPair,
string customLabel,
bool includePlaceholderTags,
bool includeTagPairs,
List <string> postCustomizationBatch,
IsoLanguage sourceLanguage,
IsoLanguage targetLanguage,
bool guidedAlignment = false)
{
this.model = model;
this.customModel = customModel;
this.modelDir = new DirectoryInfo(model.InstallDir);
this.customDir = new DirectoryInfo(this.customModel.InstallDir);
this.customSource = inputPair.Source;
this.customTarget = inputPair.Target;
this.customLabel = customLabel;
this.includePlaceholderTags = includePlaceholderTags;
this.includeTagPairs = includeTagPairs;
this.inDomainValidationSource = indomainValidPair.Source;
this.inDomainValidationTarget = indomainValidPair.Target;
this.sourceLanguage = sourceLanguage;
this.targetLanguage = targetLanguage;
this.guidedAlignment = guidedAlignment;
}
private void PreprocessInput()
{
FileInfo sourceSegModel =
this.customDir.GetFiles().Where(x => Regex.IsMatch(x.Name, "source.(spm|bpe)")).Single();
FileInfo targetSegModel =
this.customDir.GetFiles().Where(x => Regex.IsMatch(x.Name, "target.(spm|bpe)")).Single();
this.segmentationMethod = sourceSegModel.Extension;
var targetPrefix = this.model.TargetLanguages.Count > 1 ? this.targetLanguage.OriginalCode : null;
this.spSource = MarianHelper.PreprocessLanguage(
this.customSource,
this.customDir,
this.sourceLanguage.OriginalCode,
sourceSegModel,
this.includePlaceholderTags,
this.includeTagPairs,
targetPrefix);
this.spTarget = MarianHelper.PreprocessLanguage(
this.customTarget,
this.customDir,
this.targetLanguage.OriginalCode,
targetSegModel,
this.includePlaceholderTags,
this.includeTagPairs);
//concatenate the out-of-domain validation set with the in-domain validation set
ParallelFilePair tatoebaValidFileInfos =
HelperFunctions.GetTatoebaFileInfos(this.sourceLanguage.ShortestIsoCode, this.targetLanguage.ShortestIsoCode);
if (tatoebaValidFileInfos == null)
{
tatoebaValidFileInfos = HelperFunctions.GenerateDummyOODValidSet(this.customDir);
}
ParallelFilePair combinedValid = new ParallelFilePair(
tatoebaValidFileInfos,
new ParallelFilePair(this.inDomainValidationSource, this.inDomainValidationTarget),
this.customDir.FullName,
OpusCatMTEngineSettings.Default.OODValidSetSize);
this.spValidSource = MarianHelper.PreprocessLanguage(
combinedValid.Source,
this.customDir,
this.sourceLanguage.OriginalCode,
sourceSegModel,
this.includePlaceholderTags,
this.includeTagPairs,
targetPrefix);
this.spValidTarget = MarianHelper.PreprocessLanguage(
combinedValid.Target,
this.customDir,
this.targetLanguage.OriginalCode,
targetSegModel,
this.includePlaceholderTags,
this.includeTagPairs);
}
//Split a file pair into two randomly (used for separating a validation set from training set)
internal static (ParallelFilePair pair1, ParallelFilePair pair2) SplitFilePair(ParallelFilePair filePair, int pair2Size)
{
//First need to get the linecount of the file pair
var lines = 0;
using (var reader = filePair.Source.OpenText())
{
while (reader.ReadLine() != null)
{
lines++;
}
}
ParallelFilePair pair1 =
new ParallelFilePair(
$"{filePair.Source.DirectoryName}{Path.DirectorySeparatorChar}split1.{filePair.Source.Name}",
$"{filePair.Target.DirectoryName}{Path.DirectorySeparatorChar}split1.{filePair.Target.Name}");
ParallelFilePair pair2 =
new ParallelFilePair(
$"{filePair.Source.DirectoryName}{Path.DirectorySeparatorChar}split2.{filePair.Source.Name}",
$"{filePair.Target.DirectoryName}{Path.DirectorySeparatorChar}split2.{filePair.Target.Name}");
var nthLine = lines / pair2Size;
var writtenLines = 0;
string sourceLine, targetLine;
using (var sourcereader = filePair.Source.OpenText())
using (var sourcewriter1 = pair1.Source.CreateText())
using (var sourcewriter2 = pair2.Source.CreateText())
using (var targetreader = filePair.Target.OpenText())
using (var targetwriter1 = pair1.Target.CreateText())
using (var targetwriter2 = pair2.Target.CreateText())
{
while
(((sourceLine = sourcereader.ReadLine()) != null) &&
((targetLine = targetreader.ReadLine()) != null))
{
if (writtenLines % nthLine == 0)
{
sourcewriter2.WriteLine(sourceLine);
targetwriter2.WriteLine(targetLine);
}
else
{
sourcewriter1.WriteLine(sourceLine);
targetwriter1.WriteLine(targetLine);
}
writtenLines++;
}
}
return(pair1, pair2);
}
//combine two file pairs
public ParallelFilePair(ParallelFilePair pair1, ParallelFilePair pair2, string combinedPath, int pair1Lines = 1000, int pair2Lines = 1000)
{
this.Source = HelperFunctions.CombineFiles(
pair1.Source,
pair2.Source,
Path.Combine(combinedPath, "combined.source"),
pair1Lines,
pair2Lines);
this.Target = HelperFunctions.CombineFiles(
pair1.Target,
pair2.Target,
Path.Combine(combinedPath, "combined.target"),
pair1Lines,
pair2Lines);
}
private void PreprocessInput()
{
var sourceSpm = this.customDir.GetFiles("source.spm").Single();
var targetSpm = this.customDir.GetFiles("target.spm").Single();
this.spSource = MarianHelper.PreprocessLanguage(this.customSource, this.customDir, this.sourceCode, sourceSpm, this.includePlaceholderTags, this.includeTagPairs);
this.spTarget = MarianHelper.PreprocessLanguage(this.customTarget, this.customDir, this.targetCode, targetSpm, this.includePlaceholderTags, this.includeTagPairs);
//concatenate the out-of-domain validation set with the in-domain validation set
ParallelFilePair tatoebaValidFileInfos = HelperFunctions.GetTatoebaFileInfos(this.sourceCode, this.targetCode);
ParallelFilePair combinedValid = new ParallelFilePair(
tatoebaValidFileInfos,
new ParallelFilePair(this.inDomainValidationSource, this.inDomainValidationTarget),
this.customDir.FullName,
OpusCatMTEngineSettings.Default.OODValidSetSize);
this.spValidSource = MarianHelper.PreprocessLanguage(combinedValid.Source, this.customDir, this.sourceCode, sourceSpm, this.includePlaceholderTags, this.includeTagPairs);
this.spValidTarget = MarianHelper.PreprocessLanguage(combinedValid.Target, this.customDir, this.targetCode, targetSpm, this.includePlaceholderTags, this.includeTagPairs);
}
