edu.stanford.nlp.ie.machinereading

Class MachineReading

java.lang.Object

edu.stanford.nlp.ie.machinereading.MachineReading

public class MachineReading
extends java.lang.Object

Main driver for Machine Reading training, annotation, and evaluation. Does entity, relation, and event extraction for all corpora. This code has been adapted for 4 domains, all defined in the edu.stanford.nlp.ie.machinereading.domains package. For each domain, you need a properties file that is the only command line parameter for MachineReading. Minimally, for each domain you need to define a reader class that extends the GenericDataSetReader class and overrides the public Annotation read(String path) method. How to run: java edu.stanford.nlp.ie.machinereading.MachineReading --arguments propertiesfile This method creates an Annotation with additional objects per sentence: EntityMentions and RelationMentions. Using these objects, the classifiers that get called from MachineReading train entity and relation extractors. The simplest example domain currently is in edu.stanford.nlp.ie.machinereading.domains.roth, which is a simple entity and relation extraction using a dataset created by Dan Roth. The properties file for the domain is at projects/more/src/edu/stanford/nlp/ie/machinereading/domains/roth/roth.properties

Author:

David McCLosky, mrsmith, Mihai

Field Summary

Fields

Modifier and Type	Field and Description
GenericDataSetReader	auxReader
Extractor	consistencyChecker
static int	ENTITY_LEVEL
Extractor	entityExtractor
static int	EVENT_LEVEL
Extractor	eventExtractor
boolean	forceParseSentences
protected boolean	forceRetraining
GenericDataSetReader	reader
static int	RELATION_LEVEL
Extractor	relationExtractionPostProcessor
Extractor	relationExtractor

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	MachineReading()

Method Summary

Methods

Modifier and Type	Method and Description
Annotation	annotate(Annotation testing)
protected Annotation	annotate(Annotation testing, int partition)
GenericDataSetReader	getAuxReader()
Pair<Annotation,Annotation>[]	getDatasets()
java.util.Set<ResultsPrinter>	getEntityResultsPrinterSet()
Annotation[][]	getPredictions()
GenericDataSetReader	getReader()
java.util.Set<ResultsPrinter>	getRelationResultsPrinterSet()
protected Annotation	loadOrMakeSerializedSentences(java.lang.String sentencesPath, GenericDataSetReader reader, java.io.File serializedSentences) Gets the serialized sentences for a data set.
static void	main(java.lang.String[] args)
protected void	makeDataSets(Annotation training, Annotation testing, Annotation auxDataset)
static Extractor	makeEntityExtractor(java.lang.Class<? extends BasicEntityExtractor> entityExtractorClass, java.lang.String gazetteerPath)
static Extractor	makeExtractor(java.lang.Class<Extractor> extractorClass)
static MachineReading	makeMachineReading(java.lang.String[] args)
static MachineReading	makeMachineReadingForAnnotation(GenericDataSetReader reader, Extractor entityExtractor, Extractor relationExtractor, Extractor eventExtractor, Extractor consistencyChecker, Extractor relationPostProcessor, boolean testRelationsUsingPredictedEntities, boolean verbose) Creates a MR object to be used only for annotation purposes (no training) This is needed in order to integrate MachineReading with BaselineNLProcessor
static RelationFeatureFactory	makeRelationFeatureFactory(java.lang.Class<RelationFeatureFactory> relationFeatureFactoryClass, java.lang.String relationFeatureList, boolean doNotLexicalizeFirstArg)
static RelationMentionFactory	makeRelationMentionFactory(java.lang.Class<RelationMentionFactory> relationMentionFactoryClass)
protected java.util.List<java.lang.String>	printTask(java.lang.String taskName, java.util.Set<ResultsPrinter> printers, Annotation gold, Annotation pred)
java.util.List<java.lang.String>	run() Performs extraction.
protected boolean	serializedModelExists(java.lang.String prefix)
void	setAuxReader(GenericDataSetReader auxReader)
static void	setConsoleLevel(java.util.logging.Level level)
void	setDatasets(Pair<Annotation,Annotation>[] datasets)
void	setEntityResultsPrinterSet(java.util.Set<ResultsPrinter> entityResultsPrinterSet)
void	setExtractEntities(boolean extractEntities)
void	setExtractEvents(boolean extractEvents)
void	setExtractRelations(boolean extractRelations)
void	setForceParseSentences(boolean forceParseSentences)
static void	setLoggerLevel(java.util.logging.Level level)
void	setPredictions(Annotation[][] predictions)
void	setReader(GenericDataSetReader reader)
void	setRelationResultsPrinterSet(java.util.Set<ResultsPrinter> relationResultsPrinterSet)
protected void	train(Annotation training, int partition)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

reader

public GenericDataSetReader reader

auxReader

public GenericDataSetReader auxReader

entityExtractor

public Extractor entityExtractor

relationExtractor

public Extractor relationExtractor

relationExtractionPostProcessor

public Extractor relationExtractionPostProcessor

eventExtractor

public Extractor eventExtractor

consistencyChecker

public Extractor consistencyChecker

forceRetraining

protected boolean forceRetraining

forceParseSentences

public boolean forceParseSentences

ENTITY_LEVEL

public static final int ENTITY_LEVEL

See Also:

Constant Field Values

RELATION_LEVEL

public static final int RELATION_LEVEL

See Also:

Constant Field Values

EVENT_LEVEL

public static final int EVENT_LEVEL

See Also:

Constant Field Values

Constructor Detail

MachineReading

protected MachineReading()

Method Detail

main

public static void main(java.lang.String[] args)
                 throws java.lang.Exception

Throws:

java.lang.Exception

setLoggerLevel

public static void setLoggerLevel(java.util.logging.Level level)

setConsoleLevel

public static void setConsoleLevel(java.util.logging.Level level)

makeMachineReadingForAnnotation

public static MachineReading makeMachineReadingForAnnotation(GenericDataSetReader reader,
                                             Extractor entityExtractor,
                                             Extractor relationExtractor,
                                             Extractor eventExtractor,
                                             Extractor consistencyChecker,
                                             Extractor relationPostProcessor,
                                             boolean testRelationsUsingPredictedEntities,
                                             boolean verbose)

Creates a MR object to be used only for annotation purposes (no training) This is needed in order to integrate MachineReading with BaselineNLProcessor

makeMachineReading

public static MachineReading makeMachineReading(java.lang.String[] args)
                                         throws java.io.IOException

Throws:

java.io.IOException

run

public java.util.List<java.lang.String> run()
                                     throws java.lang.Exception

Performs extraction. This will train a new extraction model and evaluate the model on the test set. Depending on the MachineReading instance's parameters, it may skip training if a model already exists or skip evaluation. returns results string, can be compared in a utest

Throws:

java.lang.Exception

printTask

protected java.util.List<java.lang.String> printTask(java.lang.String taskName,
                                         java.util.Set<ResultsPrinter> printers,
                                         Annotation gold,
                                         Annotation pred)

train

protected void train(Annotation training,
         int partition)
              throws java.lang.Exception

Throws:

java.lang.Exception

annotate

public Annotation annotate(Annotation testing)

annotate

protected Annotation annotate(Annotation testing,
                  int partition)

makeExtractor

public static Extractor makeExtractor(java.lang.Class<Extractor> extractorClass)

makeDataSets

protected void makeDataSets(Annotation training,
                Annotation testing,
                Annotation auxDataset)

serializedModelExists

protected boolean serializedModelExists(java.lang.String prefix)

makeEntityExtractor

public static Extractor makeEntityExtractor(java.lang.Class<? extends BasicEntityExtractor> entityExtractorClass,
                            java.lang.String gazetteerPath)

makeRelationFeatureFactory

public static RelationFeatureFactory makeRelationFeatureFactory(java.lang.Class<RelationFeatureFactory> relationFeatureFactoryClass,
                                                java.lang.String relationFeatureList,
                                                boolean doNotLexicalizeFirstArg)

makeRelationMentionFactory

public static RelationMentionFactory makeRelationMentionFactory(java.lang.Class<RelationMentionFactory> relationMentionFactoryClass)

loadOrMakeSerializedSentences

protected Annotation loadOrMakeSerializedSentences(java.lang.String sentencesPath,
                                       GenericDataSetReader reader,
                                       java.io.File serializedSentences)
                                            throws java.io.IOException,
                                                   java.lang.ClassNotFoundException

Gets the serialized sentences for a data set. If the serialized sentences are already on disk, it loads them from there. Otherwise, the data set is read with the corpus reader and the serialized sentences are saved to disk.

Parameters:

sentencesPath - location of the raw data set

reader - the corpus reader

serializedSentences - where the serialized sentences should be stored on disk

Returns:

a list of RelationsSentences

Throws:

java.io.IOException

java.lang.ClassNotFoundException

setExtractEntities

public void setExtractEntities(boolean extractEntities)

setExtractRelations

public void setExtractRelations(boolean extractRelations)

setExtractEvents

public void setExtractEvents(boolean extractEvents)

setForceParseSentences

public void setForceParseSentences(boolean forceParseSentences)

setDatasets

public void setDatasets(Pair<Annotation,Annotation>[] datasets)

getDatasets

public Pair<Annotation,Annotation>[] getDatasets()

setPredictions

public void setPredictions(Annotation[][] predictions)

getPredictions

public Annotation[][] getPredictions()

setReader

public void setReader(GenericDataSetReader reader)

getReader

public GenericDataSetReader getReader()

setAuxReader

public void setAuxReader(GenericDataSetReader auxReader)

getAuxReader

public GenericDataSetReader getAuxReader()

setEntityResultsPrinterSet

public void setEntityResultsPrinterSet(java.util.Set<ResultsPrinter> entityResultsPrinterSet)

getEntityResultsPrinterSet

public java.util.Set<ResultsPrinter> getEntityResultsPrinterSet()

setRelationResultsPrinterSet

public void setRelationResultsPrinterSet(java.util.Set<ResultsPrinter> relationResultsPrinterSet)

getRelationResultsPrinterSet

public java.util.Set<ResultsPrinter> getRelationResultsPrinterSet()