pitt.search.semanticvectors
Class VectorSearcher

java.lang.Object
  pitt.search.semanticvectors.VectorSearcher

Direct Known Subclasses:

VectorSearcher.BalancedVectorSearcherPerm, VectorSearcher.VectorSearcherConvolutionSim, VectorSearcher.VectorSearcherCosine, VectorSearcher.VectorSearcherCosineSparse, VectorSearcher.VectorSearcherMaxSim, VectorSearcher.VectorSearcherPerm, VectorSearcher.VectorSearcherSubspaceSim, VectorSearcher.VectorSearcherTensorSim

public abstract class VectorSearcher
extends java.lang.Object

Class for searching vector stores using different scoring functions. Each VectorSearcher implements a particular scoring function which is normally query dependent, so each query needs its own VectorSearcher.

Nested Class Summary

static class	VectorSearcher.BalancedVectorSearcherPerm Class for searching a permuted vector store using cosine similarity.
static class	VectorSearcher.VectorSearcherConvolutionSim Class for searching a vector store using convolution similarity.
static class	VectorSearcher.VectorSearcherCosine Class for searching a vector store using cosine similarity.
static class	VectorSearcher.VectorSearcherCosineSparse Class for searching a vector store using sparse cosine similarity.
static class	VectorSearcher.VectorSearcherMaxSim Class for searching a vector store using minimum distance similarity.
static class	VectorSearcher.VectorSearcherPerm Class for searching a permuted vector store using cosine similarity.
static class	VectorSearcher.VectorSearcherSubspaceSim Class for searching a vector store using quantum disjunction similarity.
static class	VectorSearcher.VectorSearcherTensorSim Class for searching a vector store using tensor product similarity.

Constructor Summary

VectorSearcher(VectorStore queryVecStore, VectorStore searchVecStore, LuceneUtils luceneUtils)
Performs basic initialization; subclasses should normally call super() to use this.

Method Summary

java.util.LinkedList<SearchResult>	getNearestNeighbors(int numResults) This nearest neighbor search is implemented in the abstract VectorSearcher class itself: this enables all subclasses to reuse the search whatever scoring method they implement.
abstract float	getScore(float[] testVector) This needs to be filled in for each subclass.

Methods inherited from class java.lang.Object

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

Constructor Detail

VectorSearcher

public VectorSearcher(VectorStore queryVecStore,
                      VectorStore searchVecStore,
                      LuceneUtils luceneUtils)

Performs basic initialization; subclasses should normally call super() to use this.

Parameters:

queryVecStore - Vector store to use for query generation.

searchVecStore - The vector store to search.

luceneUtils - LuceneUtils object to use for query weighting. (May be null.)

Method Detail

getScore

public abstract float getScore(float[] testVector)

This needs to be filled in for each subclass. It takes an individual vector and assigns it a relevance score for this VectorSearcher.

getNearestNeighbors

public java.util.LinkedList<SearchResult> getNearestNeighbors(int numResults)

This nearest neighbor search is implemented in the abstract VectorSearcher class itself: this enables all subclasses to reuse the search whatever scoring method they implement. Since query expressions are built into the VectorSearcher, getNearestNeighbors no longer takes a query vector as an argument.

Parameters:

numResults - the number of results / length of the result list.