- Direct Known Subclasses:
public abstract class TFIDFSimilarity extends SimilarityImplementation of
Similaritywith the Vector Space Model.
Expert: Scoring API.
TFIDFSimilarity defines the components of Lucene scoring. Overriding computation of these components is a convenient way to alter Lucene scoring.
Suggested reading: Introduction To Information Retrieval, Chapter 6.
The following describes how Lucene scoring evolves from underlying information retrieval models to (efficient) implementation. We first brief on VSM Score, then derive from it Lucene's Conceptual Scoring Formula, from which, finally, evolves Lucene's Practical Scoring Function (the latter is connected directly with Lucene classes and methods).
In VSM, documents and queries are represented as weighted vectors in a multi-dimensional space, where each distinct index term is a dimension, and weights are Tf-idf values.
VSM does not require weights to be Tf-idf values, but Tf-idf values are believed to produce search results of high quality, and so Lucene is using Tf-idf. Tf and Idf are described in more detail below, but for now, for completion, let's just say that for given term t and document (or query) x, Tf(t,x) varies with the number of occurrences of term t in x (when one increases so does the other) and idf(t) similarly varies with the inverse of the number of index documents containing term t.
VSM score of document d for query q is the Cosine Similarity of the weighted query vectors V(q) and V(d):
formatting only formatting only cosine similarity formula cosine-similarity(q,d) = cosine similarity formula V(q) · V(d) ––––––––– |V(q)| |V(d)| VSM Score
Where V(q) · V(d) is the dot product of the weighted vectors, and |V(q)| and |V(d)| are their Euclidean norms.
Note: the above equation can be viewed as the dot product of the normalized weighted vectors, in the sense that dividing V(q) by its euclidean norm is normalizing it to a unit vector.
Lucene refines VSM score for both search quality and usability:
- Normalizing V(d) to the unit vector is known to be problematic in that it removes all document length information. For some documents removing this info is probably ok, e.g. a document made by duplicating a certain paragraph 10 times, especially if that paragraph is made of distinct terms. But for a document which contains no duplicated paragraphs, this might be wrong. To avoid this problem, a different document length normalization factor is used, which normalizes to a vector equal to or larger than the unit vector: doc-len-norm(d).
- At indexing, users can specify that certain documents are more important than others, by assigning a document boost. For this, the score of each document is also multiplied by its boost value doc-boost(d).
- Lucene is field based, hence each query term applies to a single field, document length normalization is by the length of the certain field, and in addition to document boost there are also document fields boosts.
- The same field can be added to a document during indexing several times, and so the boost of that field is the multiplication of the boosts of the separate additions (or parts) of that field within the document.
- At search time users can specify boosts to each query, sub-query, and each query term, hence the contribution of a query term to the score of a document is multiplied by the boost of that query term query-boost(q).
- A document may match a multi term query without containing all the terms of that query (this is correct for some of the queries).
Under the simplifying assumption of a single field in the index, we get Lucene's Conceptual scoring formula:
formatting only formatting only formatting only score(q,d) = query-boost(q) · Lucene conceptual scoring formula V(q) · V(d) ––––––––– |V(q)| · doc-len-norm(d) · doc-boost(d) Lucene Conceptual Scoring Formula
The conceptual formula is a simplification in the sense that (1) terms and documents are fielded and (2) boosts are usually per query term rather than per query.
We now describe how Lucene implements this conceptual scoring formula, and derive from it Lucene's Practical Scoring Function.
For efficient score computation some scoring components are computed and aggregated in advance:
- Query-boost for the query (actually for each query term) is known when search starts.
- Query Euclidean norm |V(q)| can be computed when search starts, as it is independent
of the document being scored. From search optimization perspective, it is a valid question
why bother to normalize the query at all, because all scored documents will be multiplied
by the same |V(q)|, and hence documents ranks (their order by score) will not be
affected by this normalization. There are two good reasons to keep this normalization:
- Recall that Cosine Similarity can be used find how similar two documents are. One can use Lucene for e.g. clustering, and use a document as a query to compute its similarity to other documents. In this use case it is important that the score of document d3 for query d1 is comparable to the score of document d3 for query d2. In other words, scores of a document for two distinct queries should be comparable. There are other applications that may require this. And this is exactly what normalizing the query vector V(q) provides: comparability (to a certain extent) of two or more queries.
- Document length norm doc-len-norm(d) and document boost doc-boost(d) are known at indexing time. They are computed in advance and their multiplication is saved as a single value in the index: norm(d). (In the equations below, norm(t in d) means norm(field(t) in doc d) where field(t) is the field associated with term t.)
Lucene's Practical Scoring Function is derived from the above. The color codes demonstrate how it relates to those of the conceptual formula:
formatting only formatting only Lucene conceptual scoring formula score(q,d) = ∑ ( tf(t in d) · idf(t)2 · t.getBoost() · norm(t,d) ) t in q Lucene Practical Scoring Function
- tf(t in d) correlates to the term's frequency,
defined as the number of times term t appears in the currently scored document
d. Documents that have more occurrences of a given term receive a higher score. Note
that tf(t in q) is assumed to be 1 and therefore it does not appear in this
equation, However if a query contains twice the same term, there will be two term-queries
with that same term and hence the computation would still be correct (although not very
efficient). The default computation for tf(t in d) in
term frequency computation
tf(t in d)=
- idf(t) stands for Inverse Document Frequency. This
value correlates to the inverse of docFreq (the number of documents in which the
term t appears). This means rarer terms give higher contribution to the total score.
idf(t) appears for t in both the query and the document, hence it is squared
in the equation. The default computation for idf(t) in
inverse document frequency computation
1 + log ( inverse document frequency computation docCount+1 ––––––––– docFreq+1 )
- t.getBoost() is a search time boost of term
t in the query q as specified in the query text (see query
syntax), or as set by wrapping with
BoostQuery. Notice that there is really no direct API for accessing a boost of one term in a multi term query, but rather multi terms are represented in a query as multi
TermQueryobjects, and so the boost of a term in the query is accessible by calling the sub-query
- norm(t,d) is an index-time boost factor that solely depends on the number of tokens of this field in the document, so that shorter fields contribute more to the score.
Fields Modifier and Type Field Description
discountOverlapsTrue if overlap tokens (tokens with a position of increment of zero) are discounted from the document's length.
Constructors Constructor Description
All Methods Instance Methods Abstract Methods Concrete Methods Modifier and Type Method Description
computeNorm(FieldInvertState state)Computes the normalization value for a field, given the accumulated state of term processing for this field (see
getDiscountOverlaps()Returns true if overlap tokens are discounted from the document's length.
idf(long docFreq, long docCount)Computes a score factor based on a term's document frequency (the number of documents which contain the term).
idfExplain(CollectionStatistics collectionStats, TermStatistics termStats)Computes a score factor for a simple term and returns an explanation for that score factor.
idfExplain(CollectionStatistics collectionStats, TermStatistics termStats)Computes a score factor for a phrase.
lengthNorm(int length)Compute an index-time normalization value for this field instance.
scorer(float boost, CollectionStatistics collectionStats, TermStatistics... termStats)Compute any collection-level weight (e.g.
setDiscountOverlaps(boolean v)Determines whether overlap tokens (Tokens with 0 position increment) are ignored when computing norm.
tf(float freq)Computes a score factor based on a term or phrase's frequency in a document.
public void setDiscountOverlaps(boolean v)Determines whether overlap tokens (Tokens with 0 position increment) are ignored when computing norm. By default this is true, meaning overlap tokens do not count when computing norms.
- See Also:
- WARNING: This API is experimental and might change in incompatible ways in the next release.
public boolean getDiscountOverlaps()Returns true if overlap tokens are discounted from the document's length.
- See Also:
public abstract float tf(float freq)Computes a score factor based on a term or phrase's frequency in a document. This value is multiplied by the
idf(long, long)factor for each term in the query and these products are then summed to form the initial score for a document.
Terms and phrases repeated in a document indicate the topic of the document, so implementations of this method usually return larger values when
freqis large, and smaller values when
freq- the frequency of a term within a document
- a score factor based on a term's within-document frequency
public Explanation idfExplain(CollectionStatistics collectionStats, TermStatistics termStats)Computes a score factor for a simple term and returns an explanation for that score factor.
The default implementation uses:
idf(docFreq, docCount);Note that
CollectionStatistics.docCount()is used instead of
TermStatistics.docFreq()is used, and when the latter is inaccurate, so is
CollectionStatistics.docCount(), and in the same direction. In addition,
CollectionStatistics.docCount()does not skew when fields are sparse.
collectionStats- collection-level statistics
termStats- term-level statistics for the term
- an Explain object that includes both an idf score factor and an explanation for the term.
public Explanation idfExplain(CollectionStatistics collectionStats, TermStatistics termStats)Computes a score factor for a phrase.
The default implementation sums the idf factor for each term in the phrase.
collectionStats- collection-level statistics
termStats- term-level statistics for the terms in the phrase
- an Explain object that includes both an idf score factor for the phrase and an explanation for each term.
public abstract float idf(long docFreq, long docCount)Computes a score factor based on a term's document frequency (the number of documents which contain the term). This value is multiplied by the
tf(float)factor for each term in the query and these products are then summed to form the initial score for a document.
Terms that occur in fewer documents are better indicators of topic, so implementations of this method usually return larger values for rare terms, and smaller values for common terms.
docFreq- the number of documents which contain the term
docCount- the total number of documents in the collection
- a score factor based on the term's document frequency
public abstract float lengthNorm(int length)Compute an index-time normalization value for this field instance.
length- the number of terms in the field, optionally
- a length normalization value
public final long computeNorm(FieldInvertState state)Description copied from class:
SimilarityComputes the normalization value for a field, given the accumulated state of term processing for this field (see
Matches in longer fields are less precise, so implementations of this method usually set smaller values when
state.getLength()is large, and larger values when
Note that for a given term-document frequency, greater unsigned norms must produce scores that are lower or equal, ie. for two encoded norms
Long.compareUnsigned(n1, n2) > 0then
SimScorer.score(freq, n1) <= SimScorer.score(freq, n2)for any legal
0is not a legal norm, so
1is the norm that produces the highest scores.
public final Similarity.SimScorer scorer(float boost, CollectionStatistics collectionStats, TermStatistics... termStats)Description copied from class:
SimilarityCompute any collection-level weight (e.g. IDF, average document length, etc) needed for scoring a query.
- Specified by:
boost- a multiplicative factor to apply to the produces scores
collectionStats- collection-level statistics, such as the number of tokens in the collection.
termStats- term-level statistics, such as the document frequency of a term across the collection.
- SimWeight object with the information this Similarity needs to score a query.