edu.nyu.jet.hmm

Class Retagger

java.lang.Object

edu.nyu.jet.hmm.Retagger

public class Retagger
extends Object

class Retagger provides methods for converting Penn part-of-speech tags into Jet part-of-speech tags, and using the result to filter the entries in the Jet lexicon.

Constructor Summary

Constructors

Constructor and Description
Retagger()

Method Summary

Methods

Modifier and Type	Method and Description
static boolean	compatible(String word, String pennPOS, Annotation jetDefn) returns true if Penn part-of-speech tag 'pennPOS', as a tag for 'word', is compatible with Jet word definition 'jetDefn'.
static String	jetToPtbPos(FeatureSet fs) given a FeatureSet fs for a Jet lexical constituent (with a 'cat' feature and possibly other features), return a Penn POS consistent with 'fs'.
static void	pruneConstit(Document d, Span zone) prunes constit annotations obtained from lexical look-up using Penn tags (recorded as tagger annotations).
static FeatureSet[]	ptbToJetFS(String word, String pennPOS) given an annotation based on Penn tag set, returns an array (possibly empty) of corresponding Jet FeatureSets, with one entry for each possible Jet category and attributes.

Methods inherited from class java.lang.Object

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

Constructor Detail

Retagger

public Retagger()

Method Detail

ptbToJetFS

public static FeatureSet[] ptbToJetFS(String word,
                      String pennPOS)

given an annotation based on Penn tag set, returns an array (possibly empty) of corresponding Jet FeatureSets, with one entry for each possible Jet category and attributes.

jetToPtbPos

public static String jetToPtbPos(FeatureSet fs)

given a FeatureSet fs for a Jet lexical constituent (with a 'cat' feature and possibly other features), return a Penn POS consistent with 'fs'. If several Penn POS's are consistent, one is arbitrarily returned. If none are consistent, 'null' is returned.

pruneConstit

public static void pruneConstit(Document d,
                Span zone)

prunes constit annotations obtained from lexical look-up using Penn tags (recorded as tagger annotations).

The following rules are applied at each token:
1. If there are no lexical entries (constit tags), generate constit annotations based on tagger tags.
2. If there is a lexical entry spanning more than one token, keep it.
3. If there are lexical entries, and some of them are consistent with the tagger tags, keep only the consistent entries. 4. If the tagger tag is POS (possessive), delete all lexical entries. 5. Otherwise (there are lexical entries, but none are consistent with the tagger tags) keep all lexical entries.

compatible

public static boolean compatible(String word,
                 String pennPOS,
                 Annotation jetDefn)

returns true if Penn part-of-speech tag 'pennPOS', as a tag for 'word', is compatible with Jet word definition 'jetDefn'.