--- /dev/null
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.lucene.misc;
+
+import org.apache.lucene.search.DefaultSimilarity;
+import org.apache.lucene.index.FieldInvertState;
+
+import java.util.Map;
+import java.util.HashMap;
+
+/**
+ * A similarity with a lengthNorm that provides for a "plateau" of
+ * equally good lengths, and tf helper functions.
+ *
+ * <p>
+ * For lengthNorm, A global min/max can be specified to define the
+ * plateau of lengths that should all have a norm of 1.0.
+ * Below the min, and above the max the lengthNorm drops off in a
+ * sqrt function.
+ * </p>
+ * <p>
+ * A per field min/max can be specified if different fields have
+ * different sweet spots.
+ * </p>
+ *
+ * <p>
+ * For tf, baselineTf and hyperbolicTf functions are provided, which
+ * subclasses can choose between.
+ * </p>
+ *
+ */
+public class SweetSpotSimilarity extends DefaultSimilarity {
+
+ private int ln_min = 1;
+ private int ln_max = 1;
+ private float ln_steep = 0.5f;
+
+ private Map<String,Number> ln_maxs = new HashMap<String,Number>(7);
+ private Map<String,Number> ln_mins = new HashMap<String,Number>(7);
+ private Map<String,Float> ln_steeps = new HashMap<String,Float>(7);
+ private Map<String,Boolean> ln_overlaps = new HashMap<String,Boolean>(7);
+
+ private float tf_base = 0.0f;
+ private float tf_min = 0.0f;
+
+ private float tf_hyper_min = 0.0f;
+ private float tf_hyper_max = 2.0f;
+ private double tf_hyper_base = 1.3d;
+ private float tf_hyper_xoffset = 10.0f;
+
+ public SweetSpotSimilarity() {
+ super();
+ }
+
+ /**
+ * Sets the baseline and minimum function variables for baselineTf
+ *
+ * @see #baselineTf
+ */
+ public void setBaselineTfFactors(float base, float min) {
+ tf_min = min;
+ tf_base = base;
+ }
+
+ /**
+ * Sets the function variables for the hyperbolicTf functions
+ *
+ * @param min the minimum tf value to ever be returned (default: 0.0)
+ * @param max the maximum tf value to ever be returned (default: 2.0)
+ * @param base the base value to be used in the exponential for the hyperbolic function (default: e)
+ * @param xoffset the midpoint of the hyperbolic function (default: 10.0)
+ * @see #hyperbolicTf
+ */
+ public void setHyperbolicTfFactors(float min, float max,
+ double base, float xoffset) {
+ tf_hyper_min = min;
+ tf_hyper_max = max;
+ tf_hyper_base = base;
+ tf_hyper_xoffset = xoffset;
+ }
+
+ /**
+ * Sets the default function variables used by lengthNorm when no field
+ * specific variables have been set.
+ *
+ * @see #lengthNorm
+ */
+ public void setLengthNormFactors(int min, int max, float steepness) {
+ this.ln_min = min;
+ this.ln_max = max;
+ this.ln_steep = steepness;
+ }
+
+ /**
+ * Sets the function variables used by lengthNorm for a specific named field.
+ *
+ * @param field field name
+ * @param min minimum value
+ * @param max maximum value
+ * @param steepness steepness of the curve
+ * @param discountOverlaps if true, <code>numOverlapTokens</code> will be
+ * subtracted from <code>numTokens</code>; if false then
+ * <code>numOverlapTokens</code> will be assumed to be 0 (see
+ * {@link DefaultSimilarity#computeNorm(String, FieldInvertState)} for details).
+ *
+ * @see #lengthNorm
+ */
+ public void setLengthNormFactors(String field, int min, int max,
+ float steepness, boolean discountOverlaps) {
+ ln_mins.put(field, Integer.valueOf(min));
+ ln_maxs.put(field, Integer.valueOf(max));
+ ln_steeps.put(field, Float.valueOf(steepness));
+ ln_overlaps.put(field, new Boolean(discountOverlaps));
+ }
+
+ /**
+ * Implemented as <code> state.getBoost() *
+ * lengthNorm(fieldName, numTokens) </code> where
+ * numTokens does not count overlap tokens if
+ * discountOverlaps is true by default or true for this
+ * specific field. */
+ @Override
+ public float computeNorm(String fieldName, FieldInvertState state) {
+ final int numTokens;
+ boolean overlaps = discountOverlaps;
+ if (ln_overlaps.containsKey(fieldName)) {
+ overlaps = ln_overlaps.get(fieldName).booleanValue();
+ }
+ if (overlaps)
+ numTokens = state.getLength() - state.getNumOverlap();
+ else
+ numTokens = state.getLength();
+
+ return state.getBoost() * computeLengthNorm(fieldName, numTokens);
+ }
+
+ /**
+ * Implemented as:
+ * <code>
+ * 1/sqrt( steepness * (abs(x-min) + abs(x-max) - (max-min)) + 1 )
+ * </code>.
+ *
+ * <p>
+ * This degrades to <code>1/sqrt(x)</code> when min and max are both 1 and
+ * steepness is 0.5
+ * </p>
+ *
+ * <p>
+ * :TODO: potential optimization is to just flat out return 1.0f if numTerms
+ * is between min and max.
+ * </p>
+ *
+ * @see #setLengthNormFactors
+ */
+ public float computeLengthNorm(String fieldName, int numTerms) {
+ int l = ln_min;
+ int h = ln_max;
+ float s = ln_steep;
+
+ if (ln_mins.containsKey(fieldName)) {
+ l = ln_mins.get(fieldName).intValue();
+ }
+ if (ln_maxs.containsKey(fieldName)) {
+ h = ln_maxs.get(fieldName).intValue();
+ }
+ if (ln_steeps.containsKey(fieldName)) {
+ s = ln_steeps.get(fieldName).floatValue();
+ }
+
+ return (float)
+ (1.0f /
+ Math.sqrt
+ (
+ (
+ s *
+ (float)(Math.abs(numTerms - l) + Math.abs(numTerms - h) - (h-l))
+ )
+ + 1.0f
+ )
+ );
+ }
+
+ /**
+ * Delegates to baselineTf
+ *
+ * @see #baselineTf
+ */
+ @Override
+ public float tf(int freq) {
+ return baselineTf(freq);
+ }
+
+ /**
+ * Implemented as:
+ * <code>
+ * (x <= min) ? base : sqrt(x+(base**2)-min)
+ * </code>
+ * ...but with a special case check for 0.
+ * <p>
+ * This degrates to <code>sqrt(x)</code> when min and base are both 0
+ * </p>
+ *
+ * @see #setBaselineTfFactors
+ */
+ public float baselineTf(float freq) {
+
+ if (0.0f == freq) return 0.0f;
+
+ return (freq <= tf_min)
+ ? tf_base
+ : (float)Math.sqrt(freq + (tf_base * tf_base) - tf_min);
+ }
+
+ /**
+ * Uses a hyperbolic tangent function that allows for a hard max...
+ *
+ * <code>
+ * tf(x)=min+(max-min)/2*(((base**(x-xoffset)-base**-(x-xoffset))/(base**(x-xoffset)+base**-(x-xoffset)))+1)
+ * </code>
+ *
+ * <p>
+ * This code is provided as a convenience for subclasses that want
+ * to use a hyperbolic tf function.
+ * </p>
+ *
+ * @see #setHyperbolicTfFactors
+ */
+ public float hyperbolicTf(float freq) {
+ if (0.0f == freq) return 0.0f;
+
+ final float min = tf_hyper_min;
+ final float max = tf_hyper_max;
+ final double base = tf_hyper_base;
+ final float xoffset = tf_hyper_xoffset;
+ final double x = (double)(freq - xoffset);
+
+ final float result = min +
+ (float)(
+ (max-min) / 2.0f
+ *
+ (
+ ( ( Math.pow(base,x) - Math.pow(base,-x) )
+ / ( Math.pow(base,x) + Math.pow(base,-x) )
+ )
+ + 1.0d
+ )
+ );
+
+ return Float.isNaN(result) ? max : result;
+
+ }
+
+}
fnp / pylucene.git / blobdiff