While this model is parameterless in the + * + * original article, + * information-based models (see {@link IBSimilarity}) introduced a + * multiplying factor. + * The default value for the {@code c} parameter is {@code 1}.
* @lucene.experimental */ public class NormalizationH1 extends Normalization { + private final float c; + + public NormalizationH1(float c) { + this.c = c; + } + + public NormalizationH1() { + this(1); + } + @Override public final float tfn(BasicStats stats, float tf, float len) { - return tf * stats.getAvgFieldLength() / len; + return tf * c * stats.getAvgFieldLength() / len; } @Override diff --git lucene/src/java/org/apache/lucene/search/similarities/NormalizationH2.java lucene/src/java/org/apache/lucene/search/similarities/NormalizationH2.java index ad7f965..cc2114a 100644 --- lucene/src/java/org/apache/lucene/search/similarities/NormalizationH2.java +++ lucene/src/java/org/apache/lucene/search/similarities/NormalizationH2.java @@ -22,11 +22,26 @@ import static org.apache.lucene.search.similarities.SimilarityBase.log2; * Normalization model in which the term frequency is inversely related to the * length. + *While this model is parameterless in the + * + * original article, the thesis + * introduces the parameterized variant. + * The default value for the {@code c} parameter is {@code 1}.
* @lucene.experimental */ public class NormalizationH2 extends Normalization { + private final float c; + + public NormalizationH2(float c) { + this.c = c; + } + + public NormalizationH2() { + this(1); + } + @Override public final float tfn(BasicStats stats, float tf, float len) { - return (float)(tf * log2(1 + stats.getAvgFieldLength() / len)); + return (float)(tf * log2(1 + c * stats.getAvgFieldLength() / len)); } @Override