public NERCombinerAnnotator(NERClassifierCombiner ner, bool verbose, int nThreads, long maxTime, int maxSentenceLength, bool fineGrained, bool entityMentions)
{
Verbose = verbose;
this.ner = ner;
this.maxTime = maxTime;
this.nThreads = nThreads;
this.maxSentenceLength = maxSentenceLength;
Properties nerProperties = new Properties();
nerProperties.SetProperty("ner.applyFineGrained", bool.ToString(fineGrained));
nerProperties.SetProperty("ner.buildEntityMentions", bool.ToString(entityMentions));
SetUpAdditionalRulesNER(nerProperties);
SetUpFineGrainedNER(nerProperties);
SetUpEntityMentionBuilding(nerProperties);
}
/// <exception cref="System.IO.IOException"/>
public NERCombinerAnnotator(Properties properties)
{
IList <string> models = new List <string>();
string modelNames = properties.GetProperty("ner.model");
if (modelNames == null)
{
modelNames = DefaultPaths.DefaultNerThreeclassModel + ',' + DefaultPaths.DefaultNerMucModel + ',' + DefaultPaths.DefaultNerConllModel;
}
if (!modelNames.IsEmpty())
{
Sharpen.Collections.AddAll(models, Arrays.AsList(modelNames.Split(",")));
}
if (models.IsEmpty())
{
// Allow for no real NER model - can just use numeric classifiers or SUTime.
// Have to unset ner.model, so unlikely that people got here by accident.
log.Info("WARNING: no NER models specified");
}
bool applyNumericClassifiers = PropertiesUtils.GetBool(properties, NERClassifierCombiner.ApplyNumericClassifiersProperty, NERClassifierCombiner.ApplyNumericClassifiersDefault);
bool applyRegexner = PropertiesUtils.GetBool(properties, NERClassifierCombiner.ApplyGazetteProperty, NERClassifierCombiner.ApplyGazetteDefault);
bool useSUTime = PropertiesUtils.GetBool(properties, NumberSequenceClassifier.UseSutimeProperty, NumberSequenceClassifier.UseSutimeDefault);
// option for setting doc date to be the present during each annotation
usePresentDateForDocDate = PropertiesUtils.GetBool(properties, "ner." + "usePresentDateForDocDate", false);
// option for setting doc date from a provided string
providedDocDate = PropertiesUtils.GetString(properties, "ner." + "providedDocDate", string.Empty);
Pattern p = Pattern.Compile("[0-9]{4}\\-[0-9]{2}\\-[0-9]{2}");
Matcher m = p.Matcher(providedDocDate);
if (!m.Matches())
{
providedDocDate = string.Empty;
}
NERClassifierCombiner.Language nerLanguage = NERClassifierCombiner.Language.FromString(PropertiesUtils.GetString(properties, NERClassifierCombiner.NerLanguageProperty, null), NERClassifierCombiner.NerLanguageDefault);
bool verbose = PropertiesUtils.GetBool(properties, "ner." + "verbose", false);
string[] loadPaths = Sharpen.Collections.ToArray(models, new string[models.Count]);
Properties combinerProperties = PropertiesUtils.ExtractSelectedProperties(properties, NERClassifierCombiner.DefaultPassDownProperties);
if (useSUTime)
{
// Make sure SUTime parameters are included
Properties sutimeProps = PropertiesUtils.ExtractPrefixedProperties(properties, NumberSequenceClassifier.SutimeProperty + '.', true);
PropertiesUtils.OverWriteProperties(combinerProperties, sutimeProps);
}
NERClassifierCombiner nerCombiner = new NERClassifierCombiner(applyNumericClassifiers, nerLanguage, useSUTime, applyRegexner, combinerProperties, loadPaths);
this.nThreads = PropertiesUtils.GetInt(properties, "ner.nthreads", PropertiesUtils.GetInt(properties, "nthreads", 1));
this.maxTime = PropertiesUtils.GetLong(properties, "ner.maxtime", 0);
this.maxSentenceLength = PropertiesUtils.GetInt(properties, "ner.maxlen", int.MaxValue);
this.language = LanguageInfo.GetLanguageFromString(PropertiesUtils.GetString(properties, "ner.language", "en"));
// in case of Spanish, use the Spanish number regexner annotator
if (language.Equals(LanguageInfo.HumanLanguage.Spanish))
{
Properties spanishNumberRegexNerProperties = new Properties();
spanishNumberRegexNerProperties["spanish.number.regexner.mapping"] = spanishNumberRegexRules;
spanishNumberRegexNerProperties["spanish.number.regexner.validpospattern"] = "^(NUM).*";
spanishNumberRegexNerProperties["spanish.number.regexner.ignorecase"] = "true";
spanishNumberAnnotator = new TokensRegexNERAnnotator("spanish.number.regexner", spanishNumberRegexNerProperties);
}
// set up fine grained ner
SetUpFineGrainedNER(properties);
// set up additional rules ner
SetUpAdditionalRulesNER(properties);
// set up entity mentions
SetUpEntityMentionBuilding(properties);
Verbose = verbose;
this.ner = nerCombiner;
}
public NERCombinerAnnotator(NERClassifierCombiner ner, bool verbose, int nThreads, long maxTime)
: this(ner, verbose, nThreads, maxTime, int.MaxValue)
{
}
public NERCombinerAnnotator(NERClassifierCombiner ner, bool verbose, int nThreads, long maxTime, int maxSentenceLength)
: this(ner, verbose, nThreads, maxTime, maxSentenceLength, true, true)
{
}
public NERCombinerAnnotator(NERClassifierCombiner ner, bool verbose)
: this(ner, verbose, 1, 0, int.MaxValue)
{
}