Examples of edu.ucla.sspace.dv.DependencyVectorSpace

edu.ucla.sspace.dv.DependencyVectorSpace
lpado.de/~sebastian/pub/papers/cl07_pado.pdf">here.
Sebastian Pado and Mirella Lapata. Constructing Semantic Space Models from Parsed Corpora. Proceedings of ACL-03, Sapporo. Available here.

This algorithm operates on dependency-parsed corpora. Each sentence is represented as a parse tree. When two words are connected by a path in these trees, the path is analyzed to see if it contains semantic information, e.g. the "sbj" path connecting the sentence subject to a verb would be informative. Then the words connected by the path are updated as co-occurring. The algorith has three main points of variation:

A {@link BasisFunction} that maps the the co-occurrence of a word atthe end of a path to a specific dimension. For example, a basis function may map each occurrence of a word to a single dimension, or the function might map each occurrence to a different dimension specific to how the work was related.
A {@link PathWeight} for specifying how to value the co-occurrence.For example, each occurrence may have the same value, or the weight could be based on how long is the path that connects them.
A {@link DependencyPathAcceptor} that determines which paths are to beused in counting co-occurrences. Padó and Lapata provide three templates to match against: {@link MinimumTemplateAcceptor}, {@link MediumTemplateAcceptor}, and {@link MaximumTemplateAcceptor}. Each acceptor matches the next smaller's set of paths and additional paths. See Padó and Lapata (2007) for details.

This class offers the following three parameter options for configuration.

Property: {@value #PATH_ACCEPTOR_PROPERTY} Default: {@link MinimalTemplateAcceptor}: This property sets {@link DependencyPathAcceptor} to use for validating dependency paths. If apath is rejected it will not count towards co-occurrences.
Property: {@value #DependencyPathWeight} Default: {@link FlatPathWeight}: This property sets the method by which co-occurrences are scored. Each valid path is scored by this method. By default all paths are treated the same regardless of length.
Property: {@value #BASIS_MAPPING_PROPERTY} Default: {@link WordBasedBasisMapping}: This property determine the way in which a path is mapped to a specific dimension in the vector space. By default, only words are used as dimensions; the occurrence of a word at the end of a path is treated the same regardless of the relation connect it or length of the path.

This class is thread-safe for concurrent calls of {@link #processDocument(BufferedReader) processDocument}. At any given point in processing, the {@link #getVectorFor(String) getVector} method may be usedto access the current semantics of a word. This allows callers to track incremental changes to the semantics as the corpus is processed.

The {@link #processSpace(Properties) processSpace} method for this class doesnothing. @see edu.ucla.sspace.svs.StructuredVectorSpace @see BasisFunction @see PathWeight @see DependencyPathAcceptor @author David Jurgens

     */
    protected SemanticSpace getSpace() {
        // Ensure that the configured DependencyExtactor is in place prior to
        // constructing the SVS
        setupDependencyExtractor();        
        return new DependencyVectorSpace(
                System.getProperties(), argOptions.getIntOption('l', 0));
    }

Related Classes of edu.ucla.sspace.dv.DependencyVectorSpace

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.