edu.ucla.sspace.wordsi

Class DependencyContextExtractor

java.lang.Object

edu.ucla.sspace.wordsi.DependencyContextExtractor

All Implemented Interfaces:

ContextExtractor

Direct Known Subclasses:

PseudoWordDependencyContextExtractor, SemEvalDependencyContextExtractor

public class DependencyContextExtractor
extends Object
implements ContextExtractor

This ContextExtractor reads in documents that have been dependency parsed. Contexts are defined by a FilteredDependencyIterator, which is used to traverse all possible dependency paths rooted at each word of interest in a document. Each reachable and valid DependencyPath forms a feature and is weighted by a DependencyPathWeight.

Author:

Keith Stevens

Field Summary

Fields

Modifier and Type	Field and Description
protected DependencyExtractor	extractor The DependencyExtractor used to extract parse trees from the already parsed documents
protected DependencyContextGenerator	generator The DependencyContextGenerator responsible for processing a DependencyTreeNode and turning it into a context vector.
protected boolean	readHeader If true, the first line in a dependency document will be treated as the header of the document, and not part of the parse tree.

Constructor Summary

Constructors

Constructor and Description
DependencyContextExtractor(DependencyExtractor extractor, DependencyContextGenerator generator) Creates a new DependencyContextExtractor.
DependencyContextExtractor(DependencyExtractor extractor, DependencyContextGenerator generator, boolean readHeader) Creates a new DependencyContextExtractor.

Method Summary

Methods

Modifier and Type	Method and Description
protected boolean	acceptWord(DependencyTreeNode focusNode, String contextHeader, Wordsi wordsi) Returns true if Wordsi should generate a context vector for focusWord.
protected String	getPrimaryKey(DependencyTreeNode focusNode) Returns the token for the primary key, i.e.
protected String	getSecondaryKey(DependencyTreeNode focusNode, String contextHeader) Returns the token for the secondary key.
int	getVectorLength() Returns the maximum number of dimensions used to represent any given context.
protected String	handleContextHeader(BufferedReader document) Returns the string for the context header.
void	processDocument(BufferedReader document, Wordsi wordsi) Processes the content of document and calls Wordsi.handleContextVector(java.lang.String, java.lang.String, edu.ucla.sspace.vector.SparseDoubleVector) for each context vector that can be extracted from document.

Methods inherited from class java.lang.Object

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

Field Detail

extractor

protected final DependencyExtractor extractor

The DependencyExtractor used to extract parse trees from the already parsed documents

generator

protected final DependencyContextGenerator generator

The DependencyContextGenerator responsible for processing a DependencyTreeNode and turning it into a context vector.

readHeader

protected final boolean readHeader

If true, the first line in a dependency document will be treated as the header of the document, and not part of the parse tree.

Constructor Detail

DependencyContextExtractor

public DependencyContextExtractor(DependencyExtractor extractor,
                          DependencyContextGenerator generator)

Creates a new DependencyContextExtractor.

Parameters:

extractor - The DependencyExtractor that parses the document and returns a valid dependency tree

generator - The DependencyContextGenerator used to created context vectors based on a DependencyTreeNode.

DependencyContextExtractor

public DependencyContextExtractor(DependencyExtractor extractor,
                          DependencyContextGenerator generator,
                          boolean readHeader)

Creates a new DependencyContextExtractor.

Parameters:

extractor - The DependencyExtractor that parses the document and returns a valid dependency tree

generator - The DependencyContextGenerator used to created context vectors based on a DependencyTreeNode.

readheader - If true, the first line in a dependency tree document will be discarded from the tree and used as a header.

Method Detail

getVectorLength

public int getVectorLength()

Returns the maximum number of dimensions used to represent any given context.

Specified by:

getVectorLength in interface ContextExtractor

processDocument

public void processDocument(BufferedReader document,
                   Wordsi wordsi)

Processes the content of document and calls Wordsi.handleContextVector(java.lang.String, java.lang.String, edu.ucla.sspace.vector.SparseDoubleVector) for each context vector that can be extracted from document.

Specified by:

processDocument in interface ContextExtractor

acceptWord

protected boolean acceptWord(DependencyTreeNode focusNode,
                 String contextHeader,
                 Wordsi wordsi)

Returns true if Wordsi should generate a context vector for focusWord.

getPrimaryKey

protected String getPrimaryKey(DependencyTreeNode focusNode)

Returns the token for the primary key, i.e. the focus word. This is just the text of the focusNode.

getSecondaryKey

protected String getSecondaryKey(DependencyTreeNode focusNode,
                     String contextHeader)

Returns the token for the secondary key. If a contextHeader is provided, this is the contextHeader, otherwise it is the word for the focusNode.

handleContextHeader

protected String handleContextHeader(BufferedReader document)
                              throws IOException

Returns the string for the context header. If readHeader is true, this returns the first line, otherwise it returns null.

Throws:

IOException