blob: 3136102b6a27ead4dd8dd1a11312d88687aa8ee6 [file] [log] [blame]
package joshua.decoder.ff;
import java.util.List;
import joshua.decoder.ff.state_maintenance.DPState;
import joshua.decoder.ff.tm.Rule;
import joshua.decoder.hypergraph.HGNode;
import joshua.corpus.Vocabulary;
import joshua.decoder.chart_parser.SourcePath;
/**
* This feature is fired when an out-of-vocabulary word (with respect to the translation model) is
* entered into the chart. OOVs work in the following manner: for each word in the input that is OOV
* with respect to the translation model, we create a rule that pushes that word through
* untranslated (the suffix "_OOV" can optionally be appended according to the runtime parameter
* "mark-oovs") . These rules are all stored in a grammar whose owner is "oov". The OOV feature
* function template then fires the "OOVPenalty" feature whenever it is asked to score an OOV rule.
*
* @author Matt Post <post@cs.jhu.edu>
*/
public class OOVFF extends StatelessFF {
private int ownerID = -1;
public OOVFF(FeatureVector weights) {
super(weights, "OOVPenalty");
ownerID = Vocabulary.id("oov");
}
/**
* OOV rules cover exactly one word, and such rules belong to a grammar whose owner is "oov". Each
* OOV fires the OOVPenalty feature with a value of 1, so the cost is simply the weight, which was
* cached when the feature was created.
*/
@Override
public DPState compute(Rule rule, List<HGNode> tailNodes, int i, int j, SourcePath sourcePath,
int sentID, Accumulator acc) {
if (rule != null && this.ownerID == rule.getOwner())
acc.add(name, -100.0f);
return null;
}
/**
* It's important for the OOV feature to contribute to the rule's estimated cost, so that OOV
* rules (which are added for all words, not just ones without translation options) get sorted
* to the bottom during cube pruning.
*
* Important! estimateCost returns the *weighted* feature value.
*/
@Override
public float estimateCost(Rule rule, int sentID) {
if (rule != null && this.ownerID == rule.getOwner())
return weights.get(name);
return 0.0f;
}
}