edu.stanford.nlp.naturalli

Class Util

java.lang.Object

edu.stanford.nlp.naturalli.Util

public class Util
extends java.lang.Object

TODO(gabor) JavaDoc

Author:

Gabor Angeli

Field Summary

Fields

Modifier and Type	Field and Description
static java.util.Set<java.lang.String>	PRIVATIVE_ADJECTIVES The dictionary of privative adjectives, as per http://hci.stanford.edu/cstr/reports/2014-04.pdf

Constructor Summary

Constructors

Constructor and Description
Util()

Method Summary

Modifier and Type	Method and Description
static void	annotate(CoreMap sentence, AnnotationPipeline pipeline) TODO(gabor) JavaDoc
static java.util.List<SemanticGraphEdge>	cleanTree(SemanticGraph tree, SemanticGraph originalTree) Fix some bizarre peculiarities with certain trees.
static void	dumpAccuracy(Classifier<ClauseSplitter.ClauseClassifierLabel,java.lang.String> classifier, GeneralDataset<ClauseSplitter.ClauseClassifierLabel,java.lang.String> dataset) A helper function for dumping the accuracy of the trained classifier.
static Span	extractNER(java.util.List<CoreLabel> tokens, Span seed) Returns a coherent NER span from a list of tokens.
static java.lang.String	guessNER(java.util.List<CoreLabel> tokens) TODO(gabor) JavaDoc
static java.lang.String	guessNER(java.util.List<CoreLabel> tokens, Span span) TODO(gabor) JavaDoc
static boolean	isCyclic(SemanticGraph tree) Determine if a tree is cyclic.
static boolean	isTree(SemanticGraph tree) A little utility function to make sure a SemanticGraph is a tree.
static boolean	nerOverlap(java.util.List<CoreLabel> tokens, Span a, Span b)
static boolean	nerOverlap(java.util.List<CoreLabel> tokens, Span a, Span b, java.util.Optional<SemanticGraph> parse) Returns true if the given two spans denote the same consistent NER chunk.
static void	stripPrepCases(SemanticGraph tree, SemanticGraph originalTree) Strip away case edges, if the incoming edge is a preposition.
static Span	tokensToSpan(java.util.List<? extends HasIndex> tokens) Construct the spanning span of the given list of tokens.
static boolean	verifyRemoval(SemanticGraph tree, java.util.Collection<SemanticGraphEdge> removeEdges, SemanticGraphEdge toKeep, IndexedWord dependent) Verify that when removing edges from the graph, the graph is not disconnected after the removal.

Methods inherited from class java.lang.Object

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

Field Detail

PRIVATIVE_ADJECTIVES

public static final java.util.Set<java.lang.String> PRIVATIVE_ADJECTIVES

The dictionary of privative adjectives, as per http://hci.stanford.edu/cstr/reports/2014-04.pdf

Constructor Detail

Util

public Util()

Method Detail

guessNER

public static java.lang.String guessNER(java.util.List<CoreLabel> tokens,
                                        Span span)

TODO(gabor) JavaDoc

Parameters:

tokens -

span -

Returns:

guessNER

public static java.lang.String guessNER(java.util.List<CoreLabel> tokens)

TODO(gabor) JavaDoc

Parameters:

tokens -

Returns:

extractNER

public static Span extractNER(java.util.List<CoreLabel> tokens,
                              Span seed)

Returns a coherent NER span from a list of tokens.

Parameters:

tokens - The tokens of the entire sentence.

seed - The seed span of the intended NER span that should be expanded.

Returns:

A 0 indexed span corresponding to a coherent NER chunk from the given seed.

annotate

public static void annotate(CoreMap sentence,
                            AnnotationPipeline pipeline)

TODO(gabor) JavaDoc

Parameters:

sentence -

pipeline -

verifyRemoval

public static boolean verifyRemoval(SemanticGraph tree,
                                    java.util.Collection<SemanticGraphEdge> removeEdges,
                                    SemanticGraphEdge toKeep,
                                    IndexedWord dependent)

Verify that when removing edges from the graph, the graph is not disconnected after the removal.
The edge specified in toKeep will be kept.

cleanTree

public static java.util.List<SemanticGraphEdge> cleanTree(SemanticGraph tree,
                                                          SemanticGraph originalTree)

Fix some bizarre peculiarities with certain trees. So far, these include:

• Sometimes there's a node from a word to itself. This seems wrong.

Parameters:

tree - The tree to clean (in place!).

Returns:

A list of extra edges, which are valid but were removed.

stripPrepCases

public static void stripPrepCases(SemanticGraph tree,
                                  SemanticGraph originalTree)

Strip away case edges, if the incoming edge is a preposition. This replicates the behavior of the old Stanford dependencies on universal dependencies.

Parameters:

tree - The tree to modify in place.

isCyclic

public static boolean isCyclic(SemanticGraph tree)

Determine if a tree is cyclic.

Parameters:

tree - The tree to check.

Returns:

True if the tree has at least once cycle in it.

isTree

public static boolean isTree(SemanticGraph tree)

A little utility function to make sure a SemanticGraph is a tree.

Parameters:

tree - The tree to check.

Returns:

True if this SemanticGraph is a tree (versus a DAG, or Graph).

nerOverlap

public static boolean nerOverlap(java.util.List<CoreLabel> tokens,
                                 Span a,
                                 Span b,
                                 java.util.Optional<SemanticGraph> parse)

Returns true if the given two spans denote the same consistent NER chunk. That is, if we call extractNER(List, Span) on these two spans, they would return the same span.

Parameters:

tokens - The tokens in the sentence.

a - The first span.

b - The second span.

parse - The parse tree to traverse looking for coreference chains to exploit.

Returns:

True if these two spans contain exactly the same NER.

nerOverlap

public static boolean nerOverlap(java.util.List<CoreLabel> tokens,
                                 Span a,
                                 Span b)

See Also:

nerOverlap(List, Span, Span, Optional)

dumpAccuracy

public static void dumpAccuracy(Classifier<ClauseSplitter.ClauseClassifierLabel,java.lang.String> classifier,
                                GeneralDataset<ClauseSplitter.ClauseClassifierLabel,java.lang.String> dataset)

A helper function for dumping the accuracy of the trained classifier.

Parameters:

classifier - The classifier to evaluate.

dataset - The dataset to evaluate the classifier on.

tokensToSpan

public static Span tokensToSpan(java.util.List<? extends HasIndex> tokens)

Construct the spanning span of the given list of tokens.

Parameters:

tokens - The tokens that should define the span.

Returns:

A span (0-indexed) that covers all of the tokens.