package edu.ucla.sspace.wordsi;
import edu.ucla.sspace.basis.BasisMapping;
import edu.ucla.sspace.basis.StringBasisMapping;
import edu.ucla.sspace.vector.CompactSparseVector;
import edu.ucla.sspace.vector.SparseDoubleVector;
import java.io.BufferedReader;
import java.io.IOError;
import java.io.IOException;
/**
* A {@link ContextExtractor} that assumes that the corpus has already been
* pre-processed and each document is a single line with the following format:
*
* </p>
* header score,feature(|score, feature)*
* </p>
*
* where header is some id that identifies the word being represented by this
* context, score is a double, and feature is some string. With this style of
* corpus, {@link PreComputedContextExtractor} will obtain a dimension for each
* feature and transform the line into a {@link SparseDoubleVector}, passing it
* to {@link Wordsi} for futher processing.
*
* @author Keith Stevens
*/
public class PreComputedContextExtractor implements ContextExtractor {
/**
* The {@link BasisMapping} responsible for determining feature indices.
*/
private final BasisMapping<String, String> basis;
/**
* Constructs a new {@link PreComputedContextExtractor} using a {@link
* StringBasisMapping}.
*/
public PreComputedContextExtractor() {
this(new StringBasisMapping());
}
/**
* Constructs a new {@link PreComputedContextExtractor} using the given
* {@link BasisMapping}.
*/
public PreComputedContextExtractor(BasisMapping<String, String> basis) {
this.basis = basis;
}
/**
* {@inheritDoc}
*/
public void processDocument(BufferedReader document, Wordsi wordsi) {
String line;
try {
line = document.readLine();
} catch (IOException ioe) {
throw new IOError(ioe);
}
// Split the header from the rest of the context. The header must be
// the focus word for this context.
String[] headerRest = line.split("\\s+", 2);
// Reject any words not accepted by wordsi.
if (!wordsi.acceptWord(headerRest[0]))
return;
SparseDoubleVector context = new CompactSparseVector();
// Iterate through each feature and convert it to a dimension for the
// context vector.
for (String item : headerRest[1].split("\\|")) {
String[] featureScore = item.split(",", 2);
double score = Double.parseDouble(featureScore[0]);
int dimension = basis.getDimension(featureScore[1]);
if (dimension >= 0)
context.set(dimension, score);
}
wordsi.handleContextVector(headerRest[0], headerRest[0], context);
}
/**
* {@inheritDoc}
*/
public int getVectorLength() {
return basis.numDimensions();
}
}