Class CarmelUniformTermPruningPolicy


  • public class CarmelUniformTermPruningPolicy
    extends TermPruningPolicy
    Enhanced implementation of Carmel Uniform Pruning,

    TermPositions whose in-document frequency is below a specified threshold

    See CarmelTopKTermPruningPolicy for link to the paper describing this policy. are pruned.

    Conclusions of that paper indicate that it's best to compute per-term thresholds, as we do in CarmelTopKTermPruningPolicy. However for large indexes with a large number of terms that method might be too slow, and the (enhanced) uniform approach implemented here may will be faster, although it might produce inferior search quality.

    This implementation enhances the Carmel uniform pruning approach, as it allows to specify three levels of thresholds:

    • one default threshold - globally (for terms in all fields)
    • threshold per field
    • threshold per term

    These thresholds are applied so that always the most specific one takes precedence: first a per-term threshold is used if present, then per-field threshold if present, and finally the default threshold.

    Threshold are maintained in a map, keyed by either field names or terms in field:text format. precedence of these values is the following:

    Thresholds in this method of pruning are expressed as the percentage of the top-N scoring documents per term that are retained. The list of top-N documents is established by using a regular IndexSearcher and Similarity to run a simple TermQuery.

    Smaller threshold value will produce a smaller index. See TermPruningPolicy for size vs performance considerations.

    For indexes with a large number of terms this policy might be still too slow, since it issues a term query for each term in the index. In such situations, the term frequency pruning approach in TFTermPruningPolicy will be faster, though it might produce inferior search quality.

    • Constructor Detail

      • CarmelUniformTermPruningPolicy

        public CarmelUniformTermPruningPolicy​(org.apache.lucene.index.IndexReader in,
                                              Map<String,​Integer> fieldFlags,
                                              Map<String,​Float> thresholds,
                                              float defThreshold,
                                              org.apache.lucene.search.Similarity sim)
    • Method Detail

      • pruneTermEnum

        public boolean pruneTermEnum​(org.apache.lucene.index.TermEnum te)
                              throws IOException
        Description copied from class: TermPruningPolicy
        Pruning of all postings for a term (invoked once per term).
        Specified by:
        pruneTermEnum in class TermPruningPolicy
        Parameters:
        te - positioned term enum.
        Returns:
        true if all postings for this term should be removed, false otherwise.
        Throws:
        IOException
      • initPositionsTerm

        public void initPositionsTerm​(org.apache.lucene.index.TermPositions tp,
                                      org.apache.lucene.index.Term t)
                               throws IOException
        Description copied from class: TermPruningPolicy
        Called when moving TermPositions to a new Term.
        Specified by:
        initPositionsTerm in class TermPruningPolicy
        Parameters:
        tp - input term positions
        t - current term
        Throws:
        IOException
      • pruneAllPositions

        public boolean pruneAllPositions​(org.apache.lucene.index.TermPositions termPositions,
                                         org.apache.lucene.index.Term t)
                                  throws IOException
        Description copied from class: TermPruningPolicy
        Prune all postings per term (invoked once per term per doc)
        Specified by:
        pruneAllPositions in class TermPruningPolicy
        Parameters:
        termPositions - positioned term positions. Implementations MUST NOT advance this by calling TermPositions methods that advance either the position pointer (next, skipTo) or term pointer (seek).
        t - current term
        Returns:
        true if the current posting should be removed, false otherwise.
        Throws:
        IOException
      • pruneTermVectorTerms

        public int pruneTermVectorTerms​(int docNumber,
                                        String field,
                                        String[] terms,
                                        int[] freqs,
                                        org.apache.lucene.index.TermFreqVector tfv)
                                 throws IOException
        Description copied from class: TermPruningPolicy
        Pruning of individual terms in term vectors.
        Specified by:
        pruneTermVectorTerms in class TermPruningPolicy
        Parameters:
        docNumber - document number
        field - field name
        terms - array of terms
        freqs - array of term frequencies
        tfv - the original term frequency vector
        Returns:
        0 if no terms are to be removed, positive number to indicate how many terms need to be removed. The same number of entries in the terms array must be set to null to indicate which terms to remove.
        Throws:
        IOException
      • pruneSomePositions

        public int pruneSomePositions​(int docNum,
                                      int[] positions,
                                      org.apache.lucene.index.Term curTerm)
        Description copied from class: TermPruningPolicy
        Prune some postings per term (invoked once per term per doc).
        Specified by:
        pruneSomePositions in class TermPruningPolicy
        Parameters:
        docNum - current document number
        positions - original term positions in the document (and indirectly term frequency)
        curTerm - current term
        Returns:
        0 if no postings are to be removed, or positive number to indicate how many postings need to be removed. The same number of entries in the positions array must be set to -1 to indicate which positions to remove.