it.cnr.isti.hpc.dexter

Class Dexter

java.lang.Object

it.cnr.isti.hpc.dexter.Dexter

All Implemented Interfaces:

Tagger

public class Dexter
extends Object
implements Tagger

Implements a standard entity linker, given a document the text is processed with a Spotter that returns the possible mentions detected as a SpotMatchList. Each SpotMatch contains a list of candidate entities for the spot (see EntityMatchList). The Disambiguator takes in input the SpotMatch list and takes care to select only one entity for spot (if the spot has more than one entity). The Disambiguator also gives a relevance score to each EntityMatch selected and rank the entity matches by relevance. The Disambiguator usually exploits a Relatedness function that provides the semantic distance between two entities to perform the disambiguation. Dexter takes care to remove the entity matches that overlaps in the text, removing the entity matches with the lower scores. The Dexter tagger can use external spotter/disambiguation/relatedness classes if they are set in the project.properties file, you only have to implement the correspondent interface, to produce a jar with your code and put it in the folder libs.

By default Dexter uses the DictionarySpotter as spotter, the the TopScoreEntityDisambiguator as disambiguator, and the MilneRelatedness as relatedness. If you do not set different classes in the property file this component will be used.

Author:

Diego Ceccarelli Created on Sep 5, 2012

See Also:

Tagger

Constructor Summary

Constructors

Constructor and Description
Dexter()

Method Summary

Methods

Modifier and Type	Method and Description
SpotMatchList	spot(Document doc)
String	stats()
EntityMatchList	tag(Document doc) Takes a Document performs the entity linking.

Methods inherited from class java.lang.Object

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

Constructor Detail

Dexter

public Dexter()

Method Detail

spot

public SpotMatchList spot(Document doc)

tag

public EntityMatchList tag(Document doc)

Description copied from interface: Tagger

Takes a Document performs the entity linking. It returns an {link EntityMatchList}, that maps some spots detected in the document with one (and only one) entity.

Specified by:

tag in interface Tagger

Parameters:

doc - - a document to annotate.

stats

public String stats()