edu.stanford.nlp.dcoref

Class SieveCoreferenceSystem

java.lang.Object

edu.stanford.nlp.dcoref.SieveCoreferenceSystem

public class SieveCoreferenceSystem
extends Object

Multi-pass Sieve coreference resolution system (see EMNLP 2010 paper).

The main entry point for API is coref(Document document). The output is a map from CorefChain ID to corresponding CorefChain.

Author:

Jenny Finkel, Mihai Surdeanu, Karthik Raghunathan, Heeyoung Lee, Sudarshan Rangarajan

Field Summary

Fields

Modifier and Type	Field and Description
String	conllMentionEvalScript Path for the official CoNLL scorer
static Logger	logger

Constructor Summary

Constructors

Constructor and Description
SieveCoreferenceSystem(Properties props)

Method Summary

Methods

Modifier and Type	Method and Description
static boolean	checkClusters(Logger logger, String tag, Document document)
Map<Integer,CorefChain>	coref(Document document) Extracts coreference clusters.
static void	debugPrintMentions(PrintStream out, String tag, List<List<Mention>> mentions)
Dictionaries	dictionaries()
boolean	doScore()
static List<List<Mention>>	filterMentionsWithSingletonClusters(Document document, List<List<Mention>> mentions) Remove singleton clusters
static String	formatPennTree(Tree parseTree) For printing tree in a better format
static String	getConllEvalSummary(String conllMentionEvalScript, String goldFile, String predictFile)
static List<Pair<IntTuple,IntTuple>>	getLinks(Map<Integer,CorefChain> result)
static LogisticClassifier<String,String>	getSingletonPredictorFromSerializedFile(String serializedFile)
static String	initializeAndRunCoref(Properties props) Returns the name of the log file that this method writes.
void	initScorers()
static void	main(String[] args) Needs the following properties: -props 'Location of coref.properties'
void	optimizeSieveOrdering(MentionExtractor mentionExtractor, Properties props, String timestamp) Given a set of sieves, select an optimal ordering for the sieves by iterating over sieves, and selecting the one that gives the best score and adding sieves one at a time until no more sieves left
static void	printConllOutput(Document document, PrintWriter writer, boolean gold)
static void	printConllOutput(Document document, PrintWriter writer, boolean gold, boolean filterSingletons)
static void	printConllOutput(Document document, PrintWriter writer, List<List<Mention>> orderedMentions, boolean gold)
void	printF1(boolean printF1First)
protected static void	printList(Logger logger, String... args)
static void	printRawDoc(Document document, boolean gold) Print raw document for analysis
void	printTopK(Logger logger, Document document, Semantics semantics) Print logs for error analysis
static double	runAndScoreCoref(SieveCoreferenceSystem corefSystem, MentionExtractor mentionExtractor, Properties props, String timeStamp)
static void	runAndScoreCorefDist(String runDistCmd, Properties props, String propsFile) Run and score coref distributed
static void	runConllEval(String conllMentionEvalScript, String goldFile, String predictFile, String evalFile, String errFile)
Semantics	semantics()
String	sieveClassName(int sieveIndex)
static String	signature(Properties props)

Methods inherited from class java.lang.Object

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

Field Detail

logger

public static final Logger logger

conllMentionEvalScript

public final String conllMentionEvalScript

Path for the official CoNLL scorer

Constructor Detail

SieveCoreferenceSystem

public SieveCoreferenceSystem(Properties props)
                       throws Exception

Throws:

Exception

Method Detail

signature

public static String signature(Properties props)

initScorers

public void initScorers()

doScore

public boolean doScore()

dictionaries

public Dictionaries dictionaries()

semantics

public Semantics semantics()

sieveClassName

public String sieveClassName(int sieveIndex)

main

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

Needs the following properties: -props 'Location of coref.properties'

Throws:

Exception

initializeAndRunCoref

public static String initializeAndRunCoref(Properties props)
                                    throws Exception

Returns the name of the log file that this method writes.

Throws:

Exception

runAndScoreCoref

public static double runAndScoreCoref(SieveCoreferenceSystem corefSystem,
                      MentionExtractor mentionExtractor,
                      Properties props,
                      String timeStamp)
                               throws Exception

Throws:

Exception

runAndScoreCorefDist

public static void runAndScoreCorefDist(String runDistCmd,
                        Properties props,
                        String propsFile)
                                 throws Exception

Run and score coref distributed

Throws:

Exception

optimizeSieveOrdering

public void optimizeSieveOrdering(MentionExtractor mentionExtractor,
                         Properties props,
                         String timestamp)
                           throws Exception

Given a set of sieves, select an optimal ordering for the sieves by iterating over sieves, and selecting the one that gives the best score and adding sieves one at a time until no more sieves left

Throws:

Exception

coref

public Map<Integer,CorefChain> coref(Document document)
                              throws Exception

Extracts coreference clusters. This is the main API entry point for coreference resolution. Return a map from CorefChain ID to corresponding CorefChain.

Throws:

Exception

getSingletonPredictorFromSerializedFile

public static LogisticClassifier<String,String> getSingletonPredictorFromSerializedFile(String serializedFile)

filterMentionsWithSingletonClusters

public static List<List<Mention>> filterMentionsWithSingletonClusters(Document document,
                                                      List<List<Mention>> mentions)

Remove singleton clusters

runConllEval

public static void runConllEval(String conllMentionEvalScript,
                String goldFile,
                String predictFile,
                String evalFile,
                String errFile)
                         throws IOException

Throws:

IOException

getConllEvalSummary

public static String getConllEvalSummary(String conllMentionEvalScript,
                         String goldFile,
                         String predictFile)
                                  throws IOException

Throws:

IOException

printTopK

public void printTopK(Logger logger,
             Document document,
             Semantics semantics)

Print logs for error analysis

printF1

public void printF1(boolean printF1First)

printList

protected static void printList(Logger logger,
             String... args)

formatPennTree

public static String formatPennTree(Tree parseTree)

For printing tree in a better format

printConllOutput

public static void printConllOutput(Document document,
                    PrintWriter writer,
                    boolean gold)

printConllOutput

public static void printConllOutput(Document document,
                    PrintWriter writer,
                    boolean gold,
                    boolean filterSingletons)

printConllOutput

public static void printConllOutput(Document document,
                    PrintWriter writer,
                    List<List<Mention>> orderedMentions,
                    boolean gold)

printRawDoc

public static void printRawDoc(Document document,
               boolean gold)
                        throws FileNotFoundException

Print raw document for analysis

Throws:

FileNotFoundException

getLinks

public static List<Pair<IntTuple,IntTuple>> getLinks(Map<Integer,CorefChain> result)

debugPrintMentions

public static void debugPrintMentions(PrintStream out,
                      String tag,
                      List<List<Mention>> mentions)

checkClusters

public static boolean checkClusters(Logger logger,
                    String tag,
                    Document document)