private Scorer GetScorer(Subset globalSubset, Dictionary <string, Occurences> wwoBase, string baseCleanedText, PeSScorer globalScorer)
{
var wordsInDic = globalSubset.WordsWithOccurences.Keys.Intersect(wwoBase.Keys).ToList();
var baseSubset = new Subset
{
AllWordsOccurencesSumInCorpus = globalSubset.AllWordsOccurencesSumInCorpus,
AllWordsOccurencesSumInTag = globalSubset.AllWordsOccurencesSumInTag,
WordsWithOccurences = wordsInDic.ToDictionary(w => w, w => globalSubset.WordsWithOccurences[w])
};
var baseDic = new TwisterAlgorithm(baseSubset, true, false).GetDictionary();
// kiszámoljuk a base_score-t és a global_score-t
var baseScorer = new PeSScorer(new Dictionary <int, Dictionary <string, double> > {
{ 1, baseDic }
});
var baseScore = baseScorer.GetScore(baseCleanedText, ScoreMultiplier);
if (baseScore == 0)
{
return(null);
}
var globalScore = globalScorer.GetScore(baseCleanedText, ScoreMultiplier);
if (globalScore == 0)
{
return(null);
}
return(new Scorer()
{
BaseDic = baseDic,
BaseScorer = baseScorer,
BaseScore = baseScore,
GlobalScorer = globalScorer,
GlobalScore = globalScore
});
}
public void Prepare(string processId, ClassifierSettingsElastic settings, CancellationToken token)
{
var directoryPath = GetDirectoryPath(settings.ServiceId);
try
{
var service = serviceQuery.Get(settings.ServiceId);
service.Status = (int)ServiceStatusEnum.Busy;
serviceQuery.Update(service.Id, service);
IOHelper.SafeDeleteDictionary(directoryPath, true);
var globalStoreDataSet = GlobalStore.DataSets.Get(settings.DataSetName);
var dataSet = globalStoreDataSet.DataSet;
var allDicCount = settings.NGramList.Count * settings.Tags.Count;
var counter = 0;
var lockObject = new object();
Directory.CreateDirectory(directoryPath);
foreach (var nGram in settings.NGramList)
{
var subsetCreator = new SubsetCreator(dataSet.Name, new List <string> {
DocumentElastic.TextField
}, dataSet.InterpretedFields.Select(DocumentQuery.MapDocumentObjectName).ToList(), nGram, queryFactory, globalStoreDataSet.AttachmentFields);
var actualDirectory = string.Format("{0}/{1}", directoryPath, nGram);
Directory.CreateDirectory(actualDirectory);
Parallel.ForEach(settings.Tags, parallelService.ParallelOptions(), (tag, loopState) => {
token.ThrowIfCancellationRequested();
var subset = subsetCreator.CreateByTag(tag.Id, dataSet.TagField);
var dictionary = new TwisterAlgorithm(
subset, true, false,
settings.CompressSettings.CompressCategoryOccurence,
settings.CompressSettings.CompressDataSetOccurence,
(LogicalOperatorEnum)settings.CompressSettings.CompressOperator).GetDictionary();
var dicProtoBuf = new DictionaryProtoBuf
{
Id = tag.Id,
Dictionary = dictionary,
NGram = nGram
};
lock (lockObject)
{
dicProtoBuf.Serialize(string.Format("{0}/{1}", actualDirectory, dicProtoBuf.GetFileName()));
processHandler.Changed(processId, Math.Round(++counter / (double)allDicCount * 100, 2));
}
});
}
processHandler.Finished(processId, string.Format(ServiceResources.SuccessfullyPrepared_0_Service_1, ServiceTypeEnum.Classifier, service.Name));
service.Status = (int)ServiceStatusEnum.Prepared;
serviceQuery.Update(service.Id, service);
}
catch (Exception ex)
{
var service = serviceQuery.Get(settings.ServiceId);
service.Status = (int)ServiceStatusEnum.New;
serviceQuery.Update(service.Id, service);
IOHelper.SafeDeleteDictionary(directoryPath, true);
if (ex.InnerException != null && ex.InnerException is OperationCanceledException)
{
processHandler.Cancelled(processId);
}
else
{
processHandler.Interrupted(processId, ex);
}
}
}
public void Prepare(string processId, PrcSettingsElastic settings, CancellationToken token)
{
var directoryPath = GetDirectoryPath(settings.ServiceId);
try
{
var service = serviceQuery.Get(settings.ServiceId);
service.Status = (int)ServiceStatusEnum.Busy;
serviceQuery.Update(service.Id, service);
IOHelper.SafeDeleteDictionary(directoryPath, true);
var globalStoreDataSet = GlobalStore.DataSets.Get(settings.DataSetName);
var dataSet = globalStoreDataSet.DataSet;
var progress = new Progress(settings.Tags.Count);
var subsetCreator = new SubsetCreator(dataSet.Name, new List <string> {
DocumentElastic.TextField
}, dataSet.InterpretedFields.Select(DocumentQuery.MapDocumentObjectName).ToList(), 1, queryFactory, globalStoreDataSet.AttachmentFields);
Directory.CreateDirectory(directoryPath);
var logPrefix = $"Prc Prepare {processId}";
logger.LogInformation($"{logPrefix} starts with ParallelLimit: {parallelService.ParallelLimit}, Tags Count: {settings.Tags.Count}");
var lockObject = new object();
Parallel.ForEach(settings.Tags, parallelService.ParallelOptions(), (tag, loopState) =>
{
token.ThrowIfCancellationRequested();
logger.LogInformation($"{logPrefix} preparing Tag: `{tag}`");
var subset = subsetCreator.CreateByTag(tag.Id, dataSet.TagField);
var algorithm = new TwisterAlgorithm(
subset, true, true,
settings.CompressSettings.CompressCategoryOccurence,
settings.CompressSettings.CompressDataSetOccurence,
(LogicalOperatorEnum)settings.CompressSettings.CompressOperator);
algorithm.InitTagDictionary();
var notNeededWords = subset.WordsWithOccurences.Keys.Except(
algorithm.TagDictionary
.Where(sd => sd.Value.PMI > 0)
.Select(sd => sd.Key)).ToList();
var td = algorithm.GetDictionary();
foreach (var word in notNeededWords)
{
subset.WordsWithOccurences.Remove(word);
}
lock (lockObject)
{
//dictionary serialization
var dicProtoBuf = new DictionaryProtoBuf
{
Id = tag.Id,
Dictionary = td,
NGram = 1
};
dicProtoBuf.Serialize(string.Format("{0}/{1}", directoryPath, dicProtoBuf.GetFileName()));
//subset serialization
var subsetProtoBuf = new SubsetProtoBuf
{
Id = tag.Id,
WordsWithOccurences = subset.WordsWithOccurences,
AllWordsOccurencesSumInTag = subset.AllWordsOccurencesSumInTag,
AllOccurencesSumInCorpus = subset.AllWordsOccurencesSumInCorpus
};
subsetProtoBuf.Serialize(string.Format("{0}/{1}", directoryPath, subsetProtoBuf.GetFileName()));
progress.Step();
processHandler.Changed(processId, progress.Percent.Round(2));
}
logger.LogInformation($"{logPrefix} prepared Tag: `{tag}`");
});
processHandler.Finished(processId, string.Format(ServiceResources.SuccessfullyPrepared_0_Service_1, ServiceTypeEnum.Prc, service.Name));
service.Status = (int)ServiceStatusEnum.Prepared;
serviceQuery.Update(service.Id, service);
}
catch (Exception ex)
{
var service = serviceQuery.Get(settings.ServiceId);
service.Status = (int)ServiceStatusEnum.New;
serviceQuery.Update(service.Id, service);
IOHelper.SafeDeleteDictionary(directoryPath, true);
if (ex.InnerException != null && ex.InnerException is OperationCanceledException)
{
processHandler.Cancelled(processId);
}
else
{
processHandler.Interrupted(processId, ex);
}
}
}
