beagled/Lucene.Net/Search/Similarity.cs

   1 /*
   2  * Copyright 2004 The Apache Software Foundation
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at
   7  *
   8  * http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16 using System;
  17 using Field = Lucene.Net.Documents.Field;
  18 using IndexReader = Lucene.Net.Index.IndexReader;
  19 using IndexWriter = Lucene.Net.Index.IndexWriter;
  20 using Term = Lucene.Net.Index.Term;
  21 namespace Lucene.Net.Search
  22 {
  23
  24
  25         /// <summary>Expert: Scoring API.
  26         /// <p>Subclasses implement search scoring.
  27         ///
  28         /// <p>The score of query <code>q</code> for document <code>d</code> is defined
  29         /// in terms of these methods as follows:
  30         ///
  31         /// <table cellpadding="0" cellspacing="0" border="0">
  32         /// <tr>
  33         /// <td valign="middle" align="right" rowspan="2">score(q,d) =<br></td>
  34         /// <td valign="middle" align="center">
  35         /// <big><big><big><big><big>&Sigma;</big></big></big></big></big></td>
  36         /// <td valign="middle"><small>
  37         /// {@link #Tf(int) tf}(t in d) *
  38         /// {@link #Idf(Term,Searcher) idf}(t) *
  39         /// {@link Field#getBoost getBoost}(t.Field in d) *
  40         /// {@link #LengthNorm(String,int) lengthNorm}(t.Field in d)
  41         /// </small></td>
  42         /// <td valign="middle" rowspan="2">&nbsp;*
  43         /// {@link #Coord(int,int) coord}(q,d) *
  44         /// {@link #QueryNorm(float) queryNorm}(q)
  45         /// </td>
  46         /// </tr>
  47         /// <tr>
  48         /// <td valign="top" align="right">
  49         /// <small>t in q</small>
  50         /// </td>
  51         /// </tr>
  52         /// </table>
  53         ///
  54         /// </summary>
  55         /// <seealso cref="#SetDefault(Similarity)">
  56         /// </seealso>
  57         /// <seealso cref="IndexWriter#SetSimilarity(Similarity)">
  58         /// </seealso>
  59         /// <seealso cref="Searcher#SetSimilarity(Similarity)">
  60         /// </seealso>
  61         public abstract class Similarity
  62         {
  63                 /// <summary>The Similarity implementation used by default. </summary>
  64                 private static Similarity defaultImpl = new DefaultSimilarity();
  65
  66                 /// <summary>Set the default Similarity implementation used by indexing and search
  67                 /// code.
  68                 ///
  69                 /// </summary>
  70                 /// <seealso cref="Searcher#SetSimilarity(Similarity)">
  71                 /// </seealso>
  72                 /// <seealso cref="IndexWriter#SetSimilarity(Similarity)">
  73                 /// </seealso>
  74                 public static void  SetDefault(Similarity similarity)
  75                 {
  76                         Similarity.defaultImpl = similarity;
  77                 }
  78
  79                 /// <summary>Return the default Similarity implementation used by indexing and search
  80                 /// code.
  81                 ///
  82                 /// <p>This is initially an instance of {@link DefaultSimilarity}.
  83                 ///
  84                 /// </summary>
  85                 /// <seealso cref="Searcher#SetSimilarity(Similarity)">
  86                 /// </seealso>
  87                 /// <seealso cref="IndexWriter#SetSimilarity(Similarity)">
  88                 /// </seealso>
  89                 public static Similarity GetDefault()
  90                 {
  91                         return Similarity.defaultImpl;
  92                 }
  93
  94                 /// <summary>Cache of decoded bytes. </summary>
  95                 private static readonly float[] NORM_TABLE = new float[256];
  96
  97                 /// <summary>Decodes a normalization factor stored in an index.</summary>
  98                 /// <seealso cref="#EncodeNorm(float)">
  99                 /// </seealso>
 100                 public static float DecodeNorm(byte b)
 101                 {
 102                         return NORM_TABLE[b & 0xFF];
 103                 }
 104
 105                 /// <summary>Computes the normalization value for a Field given the total number of
 106                 /// terms contained in a Field.  These values, together with Field boosts, are
 107                 /// stored in an index and multipled into scores for hits on each Field by the
 108                 /// search code.
 109                 ///
 110                 /// <p>Matches in longer fields are less precise, so implemenations of this
 111                 /// method usually return smaller values when <code>numTokens</code> is large,
 112                 /// and larger values when <code>numTokens</code> is small.
 113                 ///
 114                 /// <p>That these values are computed under {@link
 115                 /// IndexWriter#AddDocument(Document)} and stored then using
 116                 /// {#encodeNorm(float)}.  Thus they have limited precision, and documents
 117                 /// must be re-indexed if this method is altered.
 118                 ///
 119                 /// </summary>
 120                 /// <param name="fieldName">the name of the Field
 121                 /// </param>
 122                 /// <param name="numTokens">the total number of tokens contained in fields named
 123                 /// <i>fieldName</i> of <i>doc</i>.
 124                 /// </param>
 125                 /// <returns> a normalization factor for hits on this Field of this document
 126                 ///
 127                 /// </returns>
 128                 /// <seealso cref="Field#SetBoost(float)">
 129                 /// </seealso>
 130                 public abstract float LengthNorm(System.String fieldName, int numTokens);
 131
 132                 /// <summary>Computes the normalization value for a query given the sum of the squared
 133                 /// weights of each of the query terms.  This value is then multipled into the
 134                 /// weight of each query term.
 135                 ///
 136                 /// <p>This does not affect ranking, but rather just attempts to make scores
 137                 /// from different queries comparable.
 138                 ///
 139                 /// </summary>
 140                 /// <param name="sumOfSquaredWeights">the sum of the squares of query term weights
 141                 /// </param>
 142                 /// <returns> a normalization factor for query weights
 143                 /// </returns>
 144                 public abstract float QueryNorm(float sumOfSquaredWeights);
 145
 146                 /// <summary>Encodes a normalization factor for storage in an index.
 147                 ///
 148                 /// <p>The encoding uses a five-bit exponent and three-bit mantissa, thus
 149                 /// representing values from around 7x10^9 to 2x10^-9 with about one
 150                 /// significant decimal digit of accuracy.  Zero is also represented.
 151                 /// Negative numbers are rounded up to zero.  Values too large to represent
 152                 /// are rounded down to the largest representable value.  Positive values too
 153                 /// small to represent are rounded up to the smallest positive representable
 154                 /// value.
 155                 ///
 156                 /// </summary>
 157                 /// <seealso cref="Field#SetBoost(float)">
 158                 /// </seealso>
 159                 public static byte EncodeNorm(float f)
 160                 {
 161                         return FloatToByte(f);
 162                 }
 163
 164                 private static float ByteToFloat(byte b)
 165                 {
 166                         if (b == 0)
 167                         // zero is a special case
 168                                 return 0.0f;
 169                         int mantissa = b & 7;
 170                         int exponent = (b >> 3) & 31;
 171                         int bits = ((exponent + (63 - 15)) << 24) | (mantissa << 21);
 172             return BitConverter.ToSingle(BitConverter.GetBytes(bits), 0);
 173                 }
 174
 175                 private static byte FloatToByte(float f)
 176                 {
 177                         if (f < 0.0f)
 178                         // round negatives up to zero
 179                                 f = 0.0f;
 180
 181                         if (f == 0.0f)
 182                         // zero is a special case
 183                                 return 0;
 184
 185                         int bits = BitConverter.ToInt32(BitConverter.GetBytes(f), 0); // parse float into parts
 186                         int mantissa = (bits & 0xffffff) >> 21;
 187                         int exponent = (((bits >> 24) & 0x7f) - 63) + 15;
 188
 189                         if (exponent > 31)
 190                         {
 191                                 // overflow: use max value
 192                                 exponent = 31;
 193                                 mantissa = 7;
 194                         }
 195
 196                         if (exponent < 0)
 197                         {
 198                                 // underflow: use min value
 199                                 exponent = 0;
 200                                 mantissa = 1;
 201                         }
 202
 203                         return (byte) ((exponent << 3) | mantissa); // pack into a byte
 204                 }
 205
 206
 207                 /// <summary>Computes a score factor based on a term or phrase's frequency in a
 208                 /// document.  This value is multiplied by the {@link #Idf(Term, Searcher)}
 209                 /// factor for each term in the query and these products are then summed to
 210                 /// form the initial score for a document.
 211                 ///
 212                 /// <p>Terms and phrases repeated in a document indicate the topic of the
 213                 /// document, so implementations of this method usually return larger values
 214                 /// when <code>freq</code> is large, and smaller values when <code>freq</code>
 215                 /// is small.
 216                 ///
 217                 /// <p>The default implementation calls {@link #Tf(float)}.
 218                 ///
 219                 /// </summary>
 220                 /// <param name="freq">the frequency of a term within a document
 221                 /// </param>
 222                 /// <returns> a score factor based on a term's within-document frequency
 223                 /// </returns>
 224                 public virtual float Tf(int freq)
 225                 {
 226                         return Tf((float) freq);
 227                 }
 228
 229                 /// <summary>Computes the amount of a sloppy phrase match, based on an edit distance.
 230                 /// This value is summed for each sloppy phrase match in a document to form
 231                 /// the frequency that is passed to {@link #Tf(float)}.
 232                 ///
 233                 /// <p>A phrase match with a small edit distance to a document passage more
 234                 /// closely matches the document, so implementations of this method usually
 235                 /// return larger values when the edit distance is small and smaller values
 236                 /// when it is large.
 237                 ///
 238                 /// </summary>
 239                 /// <seealso cref="PhraseQuery#SetSlop(int)">
 240                 /// </seealso>
 241                 /// <param name="distance">the edit distance of this sloppy phrase match
 242                 /// </param>
 243                 /// <returns> the frequency increment for this match
 244                 /// </returns>
 245                 public abstract float SloppyFreq(int distance);
 246
 247                 /// <summary>Computes a score factor based on a term or phrase's frequency in a
 248                 /// document.  This value is multiplied by the {@link #Idf(Term, Searcher)}
 249                 /// factor for each term in the query and these products are then summed to
 250                 /// form the initial score for a document.
 251                 ///
 252                 /// <p>Terms and phrases repeated in a document indicate the topic of the
 253                 /// document, so implemenations of this method usually return larger values
 254                 /// when <code>freq</code> is large, and smaller values when <code>freq</code>
 255                 /// is small.
 256                 ///
 257                 /// </summary>
 258                 /// <param name="freq">the frequency of a term within a document
 259                 /// </param>
 260                 /// <returns> a score factor based on a term's within-document frequency
 261                 /// </returns>
 262                 public abstract float Tf(float freq);
 263
 264                 /// <summary>Computes a score factor for a simple term.
 265                 ///
 266                 /// <p>The default implementation is:<pre>
 267                 /// return idf(searcher.docFreq(term), searcher.maxDoc());
 268                 /// </pre>
 269                 ///
 270                 /// Note that {@link Searcher#MaxDoc()} is used instead of
 271                 /// {@link IndexReader#NumDocs()} because it is proportional to
 272                 /// {@link Searcher#DocFreq(Term)} , i.e., when one is inaccurate,
 273                 /// so is the other, and in the same direction.
 274                 ///
 275                 /// </summary>
 276                 /// <param name="term">the term in question
 277                 /// </param>
 278                 /// <param name="searcher">the document collection being searched
 279                 /// </param>
 280                 /// <returns> a score factor for the term
 281                 /// </returns>
 282                 public virtual float Idf(Term term, Searcher searcher)
 283                 {
 284                         return Idf(searcher.DocFreq(term), searcher.MaxDoc());
 285                 }
 286
 287                 /// <summary>Computes a score factor for a phrase.
 288                 ///
 289                 /// <p>The default implementation sums the {@link #Idf(Term,Searcher)} factor
 290                 /// for each term in the phrase.
 291                 ///
 292                 /// </summary>
 293                 /// <param name="terms">the terms in the phrase
 294                 /// </param>
 295                 /// <param name="searcher">the document collection being searched
 296                 /// </param>
 297                 /// <returns> a score factor for the phrase
 298                 /// </returns>
 299                 public virtual float Idf(System.Collections.ICollection terms, Searcher searcher)
 300                 {
 301                         float idf = 0.0f;
 302                         System.Collections.IEnumerator i = terms.GetEnumerator();
 303                         while (i.MoveNext())
 304                         {
 305                                 idf += Idf((Term) i.Current, searcher);
 306                         }
 307                         return idf;
 308                 }
 309
 310                 /// <summary>Computes a score factor based on a term's document frequency (the number
 311                 /// of documents which contain the term).  This value is multiplied by the
 312                 /// {@link #Tf(int)} factor for each term in the query and these products are
 313                 /// then summed to form the initial score for a document.
 314                 ///
 315                 /// <p>Terms that occur in fewer documents are better indicators of topic, so
 316                 /// implemenations of this method usually return larger values for rare terms,
 317                 /// and smaller values for common terms.
 318                 ///
 319                 /// </summary>
 320                 /// <param name="docFreq">the number of documents which contain the term
 321                 /// </param>
 322                 /// <param name="numDocs">the total number of documents in the collection
 323                 /// </param>
 324                 /// <returns> a score factor based on the term's document frequency
 325                 /// </returns>
 326                 public abstract float Idf(int docFreq, int numDocs);
 327
 328                 /// <summary>Computes a score factor based on the fraction of all query terms that a
 329                 /// document contains.  This value is multiplied into scores.
 330                 ///
 331                 /// <p>The presence of a large portion of the query terms indicates a better
 332                 /// match with the query, so implemenations of this method usually return
 333                 /// larger values when the ratio between these parameters is large and smaller
 334                 /// values when the ratio between them is small.
 335                 ///
 336                 /// </summary>
 337                 /// <param name="overlap">the number of query terms matched in the document
 338                 /// </param>
 339                 /// <param name="maxOverlap">the total number of terms in the query
 340                 /// </param>
 341                 /// <returns> a score factor based on term overlap with the query
 342                 /// </returns>
 343                 public abstract float Coord(int overlap, int maxOverlap);
 344                 static Similarity()
 345                 {
 346                         {
 347                                 for (int i = 0; i < 256; i++)
 348                                         NORM_TABLE[i] = ByteToFloat((byte) i);
 349                         }
 350                 }
 351         }
 352 }