eu.fbk.dkm.pikes.resources

Class NAFFilter.Builder

java.lang.Object

eu.fbk.dkm.pikes.resources.NAFFilter.Builder

Enclosing class:

NAFFilter

public static final class NAFFilter.Builder
extends Object

Configurable builder object for the creation of NAFFilters.

Supported properties accepted by withProperties(Map, String) and corresponding setter methods:

Property	Values	Corresponding method	Default
termSenseFiltering	true, false	withTermSenseFiltering(Boolean)	true
termSenseCompletion	true, false	withTermSenseCompletion(Boolean)	true
entityRemoveOverlaps	true, false	withEntityRemoveOverlaps(Boolean)	true
entitySpanFixing	true, false	withEntitySpanFixing(Boolean)	true
entityAddition	true, false	withEntityAddition(Boolean)	true
entityValueNormalization	true, false	withEntityValueNormalization(Boolean)	true
linkingCompletion	true, false	withLinkingCompletion(Boolean)	true
linkingFixing	true, false	withLinkingFixing(Boolean)	false
corefForRoleDependencies	true, false	withCorefForRoleDependencies(Boolean)	false
corefSpanFixing	true, false	withCorefSpanFixing(Boolean)	false
srlRemoveWrongRefs	true, false	withSRLRemoveWrongRefs(Boolean)	true
srlRemoveUnknownPredicates	true, false	withSRLRemoveUnknownPredicates(Boolean)	false
srlPredicateAddition	true, false	withSRLPredicateAddition(Boolean)	true
srlSelfArgFixing	true, false	withSRLSelfArgFixing(Boolean)	true
srlSenseMapping	true, false	withSRLSenseMapping(Boolean)	false
srlFrameBaseMapping	true, false	withSRLFrameBaseMapping(Boolean)	true
srlRoleLinking	none, exact, coref	withSRLRoleLinking(Boolean, Boolean)	coref (= true, true)
opinionLinking	none, exact, coref	withOpinionLinking(Boolean, Boolean)	coref (= true, true)

Method Summary

Modifier and Type	Method and Description
NAFFilter	build() Creates a NAFFilter based on the flags specified on this builder object.
NAFFilter.Builder	withCorefForRoleDependencies(Boolean corefForRoleDependencies) Specifies whether new coreference relations should be added for APPO/NMOD/TITLE edges in the dependency tree between proper nouns and role nouns.
NAFFilter.Builder	withCorefSpanFixing(Boolean corefSpanFixing) Specifies whether spans of existing coreference sets should be checked and possibly shrinked or removed.
NAFFilter.Builder	withEntityAddition(Boolean entityAddition) Specifies whether new entities should be added to the document for noun phrases not already marked as entities.
NAFFilter.Builder	withEntityRemoveOverlaps(Boolean entityRemoveOverlaps) Specifies whether entities overlapping with timex or (larger) entities should be removed.
NAFFilter.Builder	withEntitySpanFixing(Boolean entitySpanFixing) Specifies whether the spans of entities should be checked and possibly fixed, removing determiners and non-alphanumeric terms.
NAFFilter.Builder	withEntityValueNormalization(Boolean entityValueNormalization) Specifies whether normalization of numerical entity values (ordinal, cardinal, percent, money) should take place.
NAFFilter.Builder	withLinkingCompletion(Boolean linkingCompletion) Specifies whether entity links in the LinkedEntities layer should be applied to entities and predicates where missing, thus performing a kind of linking completion.
NAFFilter.Builder	withLinkingFixing(Boolean linkingFixing) Specifies whether removal of inaccurate entity links to DBpedia should occur.
NAFFilter.Builder	withOpinionLinking(Boolean opinionLinking, Boolean opinionLinkingUsingCoref) Specifies whether ExternalRef tags should be added to opinion expressions, holder and targets to lthe entities, timex and predicates their heads correspond to.
NAFFilter.Builder	withProperties(Map<?,?> properties, String prefix) Sets all the properties in the map supplied, matching an optional prefix.
NAFFilter.Builder	withSRLFrameBaseMapping(Boolean srlFrameBaseMapping) Specifies whether mapping of rolesets / roles in the SRL layer to FrameBase classes / properties should take place.
NAFFilter.Builder	withSRLPredicateAddition(Boolean srlPredicateAddition) Specifies whether new predicates can be added for verbs, noun and adjectives having exactly one sense in PropBank or NomBank but not marked in the text.
NAFFilter.Builder	withSRLPreMOnIRIs(Boolean srlPreMOnIRIs) Specifies replace reference of predicate models in NAF with premon IRIs
NAFFilter.Builder	withSRLPreprocess(Boolean srlPreprocess, Boolean srlEnableMate, Boolean srlEnableSemafor) Specifies whether to preprocess SRL layer, enabling Mate and/or Semafor outputs.
NAFFilter.Builder	withSRLRemoveUnknownPredicates(Boolean srlRemoveUnknownPredicates) Specifies whether SRL predicates with unknown PropBank/NomBank rolesets/roles in the NAF should be removed.
NAFFilter.Builder	withSRLRemoveWrongRefs(Boolean srlRemoveWrongRefs) Specifies whether ExternalRefs with wrong PropBank/NomBank rolesets/roles in the NAF should be removed.
NAFFilter.Builder	withSRLRoleLinking(Boolean srlRoleLinking, Boolean useCoref) Specifies whether ExternalRef tags should be added to SRL roles to link them to the entities, timex and predicates in the NAF the role corresponds to.
NAFFilter.Builder	withSRLSelfArgFixing(Boolean srlSelfArgFixing) Specifies whether 'self-roles' can be added for predicates where missing or removed where wrongly added.
NAFFilter.Builder	withSRLSenseMapping(Boolean srlSenseMapping) Specifies whether mapping of roleset / roles in the SRL layer should take place.
NAFFilter.Builder	withTermSenseCompletion(Boolean termSenseCompletion) Specifies whether missing term senses (BBN, SST, WN Synset, SUMO mapping ) should be completed by applying sense mappings.
NAFFilter.Builder	withTermSenseFiltering(Boolean termSenseFiltering) Specifies whether term senses (BBN, SST, WN Synset, SUMO mapping, YAGO) for proper names should be removed.

Methods inherited from class java.lang.Object

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

Method Detail

withProperties

public NAFFilter.Builder withProperties(Map<?,?> properties,
                                        @Nullable
                                        String prefix)

Sets all the properties in the map supplied, matching an optional prefix.

Parameters:

properties - the properties to configure, not null

prefix - an optional prefix used to select the relevant properties in the map

Returns:

this builder object, for call chaining

withTermSenseFiltering

public NAFFilter.Builder withTermSenseFiltering(@Nullable
                                                Boolean termSenseFiltering)

Specifies whether term senses (BBN, SST, WN Synset, SUMO mapping, YAGO) for proper names should be removed.

Parameters:

termSenseFiltering - true to enable term sense filtering, null to use default value

Returns:

this builder object, for call chaining

withTermSenseCompletion

public NAFFilter.Builder withTermSenseCompletion(@Nullable
                                                 Boolean termSenseCompletion)

Specifies whether missing term senses (BBN, SST, WN Synset, SUMO mapping ) should be completed by applying sense mappings.

Parameters:

termSenseCompletion - true to enable term sense completion, null to use default value

Returns:

this builder object, for call chaining

withEntityRemoveOverlaps

public NAFFilter.Builder withEntityRemoveOverlaps(@Nullable
                                                  Boolean entityRemoveOverlaps)

Specifies whether entities overlapping with timex or (larger) entities should be removed.

Parameters:

entityRemoveOverlaps - true, to enable removal of entities that overlap with other entities or timex; null to use the default setting

Returns:

this builder object for call chaining

withEntitySpanFixing

public NAFFilter.Builder withEntitySpanFixing(@Nullable
                                              Boolean entitySpanFixing)

Specifies whether the spans of entities should be checked and possibly fixed, removing determiners and non-alphanumeric terms. If enabled and no terms remain after fixing the span of an entity, that entity is removed.

Parameters:

entitySpanFixing - true, to enable fixing of entity spans (and possible removal of invalid entities); null to use the default setting

Returns:

this builder object, for call chaining

withEntityAddition

public NAFFilter.Builder withEntityAddition(@Nullable
                                            Boolean entityAddition)

Specifies whether new entities should be added to the document for noun phrases not already marked as entities.

Parameters:

entityAddition - true, to enable entity addition; null, to use the default setting

Returns:

this builder object, for call chaining

withEntityValueNormalization

public NAFFilter.Builder withEntityValueNormalization(@Nullable
                                                      Boolean entityValueNormalization)

Specifies whether normalization of numerical entity values (ordinal, cardinal, percent, money) should take place.

Parameters:

entityValueNormalization - true, to enable entity value normalization; null, to use the default setting

Returns:

this builder object, for call chaining

withLinkingCompletion

public NAFFilter.Builder withLinkingCompletion(@Nullable
                                               Boolean linkingCompletion)

Specifies whether entity links in the LinkedEntities layer should be applied to entities and predicates where missing, thus performing a kind of linking completion.

Parameters:

linkingCompletion - true, to perform linking completion

Returns:

this builder object, for call chaining

withLinkingFixing

public NAFFilter.Builder withLinkingFixing(@Nullable
                                           Boolean linkingFixing)

Specifies whether removal of inaccurate entity links to DBpedia should occur. If enabled, links for entities whose span is part of a stop word list are removed. The stop word list contains (multi-)words that are known to be ambiguous from an analysis of Wikipedia data.

Parameters:

linkingFixing - true to enable linking fixing; null, to use the default setting

Returns:

this builder object, for call chaining

withCorefForRoleDependencies

public NAFFilter.Builder withCorefForRoleDependencies(@Nullable
                                                      Boolean corefForRoleDependencies)

Specifies whether new coreference relations should be added for APPO/NMOD/TITLE edges in the dependency tree between proper nouns and role nouns.

Parameters:

corefForRoleDependencies - true to enable addition of coreference relations for APPO/NMOD/TITLE edges; null, to use the default setting

Returns:

this builder object, for call chaining

withCorefSpanFixing

public NAFFilter.Builder withCorefSpanFixing(@Nullable
                                             Boolean corefSpanFixing)

Specifies whether spans of existing coreference sets should be checked and possibly shrinked or removed. The following rules are applied:

• remove spans without a well-defined head in the dependency tree;
• remove spans that enclose another span in the coreference set;
• remove spans with non NNP head corresponding to a verb or to a NomBank predicate that never admit itself as a role (e.g., 'war' but not 'president'), if no span with a sumo:Process head (= event) is part of the coreference set;
• shrink spans with non NNP head that contain some NNP token, if a span with NNP head is part of the coreference set;

If a coreference set becomes empty as a result of the above filtering, it is removed from the NAF document.

Parameters:

corefSpanFixing - true to enable coreference span fixing; null to use default setting

Returns:

this builder object, for call chaining

withSRLPreprocess

public NAFFilter.Builder withSRLPreprocess(@Nullable
                                           Boolean srlPreprocess,
                                           @Nullable
                                           Boolean srlEnableMate,
                                           @Nullable
                                           Boolean srlEnableSemafor)

Specifies whether to preprocess SRL layer, enabling Mate and/or Semafor outputs. If both tools are enabled, they are combined in such a way that semafor takes precedence in case two predicates refer to the same token.

Parameters:

srlPreprocess - true, to enable preprocessing of SRL layer

srlEnableMate - true, to enable Mate output

srlEnableSemafor - true, to enable Semafor output

Returns:

this builder object, for call chaining

withSRLRemoveWrongRefs

public NAFFilter.Builder withSRLRemoveWrongRefs(@Nullable
                                                Boolean srlRemoveWrongRefs)

Specifies whether ExternalRefs with wrong PropBank/NomBank rolesets/roles in the NAF should be removed. A roleset/role is considered wrong if its lemma differs from the one of the predicate in the text (errors can arise from 'excessive' mappings, e.g. in the predicate matrix).

Parameters:

srlRemoveWrongRefs - true, if removal of ExternalRefs with wrong PB/NB rolesets/roles has to be enabled

Returns:

this builder object, for call chaining

withSRLRemoveUnknownPredicates

public NAFFilter.Builder withSRLRemoveUnknownPredicates(@Nullable
                                                        Boolean srlRemoveUnknownPredicates)

Specifies whether SRL predicates with unknown PropBank/NomBank rolesets/roles in the NAF should be removed. A roleset/role is wrong if it does not appear in PropBank/NomBank frame files (SRL tools such as Mate may detect predicates for unknown rolesets, to increase recall).

Parameters:

srlRemoveUnknownPredicates - true, if removal of predicates with unknown PB/NB rolesets/roles has to be enabled

Returns:

this builder object, for call chaining

withSRLPredicateAddition

public NAFFilter.Builder withSRLPredicateAddition(@Nullable
                                                  Boolean srlPredicateAddition)

Specifies whether new predicates can be added for verbs, noun and adjectives having exactly one sense in PropBank or NomBank but not marked in the text.

Parameters:

srlPredicateAddition - true, to enable predicate addition; null to use the default setting

Returns:

this builder object, for call chaining

withSRLSelfArgFixing

public NAFFilter.Builder withSRLSelfArgFixing(@Nullable
                                              Boolean srlSelfArgFixing)

Specifies whether 'self-roles' can be added for predicates where missing or removed where wrongly added. If set, for each recognized predicate the filter checks whether the predicate term has also been marked as role. IF it is not marked in the NAF but it is always marked in NomBank training set THEN the filter adds a new role for the predicate term, using the semantic role in NomBank training set. If already marked whereas no marking should happen based on previous criteria, then the role is removed.

Parameters:

srlSelfArgFixing - true if role addition is enabled

Returns:

this builder object, for call chaining

withSRLSenseMapping

public NAFFilter.Builder withSRLSenseMapping(@Nullable
                                             Boolean srlSenseMapping)

Specifies whether mapping of roleset / roles in the SRL layer should take place. If enabled, new external refs are added to map NomBank rolesets and roles to PropBank and to map PropBank rolesets and roles to VerbNet and FrameNet, based on the predicate matrix.

Parameters:

srlSenseMapping - true, to enable SRL sense mapping; null, to use the default setting

Returns:

this builder object, for call chaining

withSRLFrameBaseMapping

public NAFFilter.Builder withSRLFrameBaseMapping(@Nullable
                                                 Boolean srlFrameBaseMapping)

Specifies whether mapping of rolesets / roles in the SRL layer to FrameBase classes / properties should take place. If enabled, new external refs for FrameBase targets are added where possible.

Parameters:

srlFrameBaseMapping - true, to enable SRL to FrameBase mapping; null, to use the default setting

Returns:

this builder object, for call chaining

withSRLRoleLinking

public NAFFilter.Builder withSRLRoleLinking(@Nullable
                                            Boolean srlRoleLinking,
                                            @Nullable
                                            Boolean useCoref)

Specifies whether ExternalRef tags should be added to SRL roles to link them to the entities, timex and predicates in the NAF the role corresponds to. The correspondence between a role and entities/predicates is computed based on the evaluation of regular expressions on the dependency tree that take properly into account coordination and prepositions (e.g., in 'Tom speaks to Al, John and the friend of Jack', the A1 role 'to Al, John and the friend of Jack' is linked to the entities 'Al' and 'John' but not 'Jack'). If useCoref is specified, SRL roles are also linked to entities, timex and predicates reachable via coreference chains.

Parameters:

srlRoleLinking - true, to enable this filtering; null, to use the default setting

useCoref - true, to enable linking to coreferring entities/timex/predicates; null, to use the default setting

Returns:

this builder object, for call chaining

withSRLPreMOnIRIs

public NAFFilter.Builder withSRLPreMOnIRIs(@Nullable
                                           Boolean srlPreMOnIRIs)

Specifies replace reference of predicate models in NAF with premon IRIs

Parameters:

srlPreMOnIRIs - true to enable IRI replacement, null to use default value

Returns:

this builder object, for call chaining

withOpinionLinking

public NAFFilter.Builder withOpinionLinking(@Nullable
                                            Boolean opinionLinking,
                                            @Nullable
                                            Boolean opinionLinkingUsingCoref)

Specifies whether ExternalRef tags should be added to opinion expressions, holder and targets to lthe entities, timex and predicates their heads correspond to.

Parameters:

opinionLinking - true, to enable this linking; null, to use the default setting

opinionLinkingUsingCoref - true, to enable linking to coreferring entities/timex/predicates; null, to use the default setting

Returns:

this builder object, for call chaining

build

public NAFFilter build()

Creates a NAFFilter based on the flags specified on this builder object.

Returns:

the constructed NAFFilter