Apache Solr Documentation

6.4 Ref Guide (PDF Download)
Solr Tutorial
Solr Community Wiki

Older Versions of this Guide (PDF)

6.5 Draft Ref Guide Topics

Meta-Documentation

This Unreleased Guide Will Cover Apache Solr 6.5

Skip to end of metadata
Go to start of metadata

You configure each filter with a <filter> element in schema.xml as a child of <analyzer>, following the <tokenizer> element. Filter definitions should follow a tokenizer or another filter definition because they take a TokenStream as input. For example.

The class attribute names a factory class that will instantiate a filter object as needed. Filter factory classes must implement the org.apache.solr.analysis.TokenFilterFactory interface. Like tokenizers, filters are also instances of TokenStream and thus are producers of tokens. Unlike tokenizers, filters also consume tokens from a TokenStream. This allows you to mix and match filters, in any order you prefer, downstream of a tokenizer.

Arguments may be passed to tokenizer factories to modify their behavior by setting attributes on the <filter> element. For example:

The following sections describe the filter factories that are included in this release of Solr.

For user tips about Solr's filters, see http://wiki.apache.org/solr/AnalyzersTokenizersTokenFilters.

Filters discussed in this section:

ASCII Folding Filter

This filter converts alphabetic, numeric, and symbolic Unicode characters which are not in the Basic Latin Unicode block (the first 127 ASCII characters) to their ASCII equivalents, if one exists. This filter converts characters from the following Unicode blocks:

Factory class: solr.ASCIIFoldingFilterFactory

Arguments: None

Example:

In: "á" (Unicode character 00E1)

Out: "a" (ASCII character 97)

Beider-Morse Filter

Implements the Beider-Morse Phonetic Matching (BMPM) algorithm, which allows identification of similar names, even if they are spelled differently or in different languages. More information about how this works is available in the section on Phonetic Matching.

BeiderMorseFilter changed its behavior in Solr 5.0 due to an update to version 3.04 of the BMPM algorithm. Older version of Solr implemented BMPM version 3.00 (see http://stevemorse.org/phoneticinfo.htm). Any index built using this filter with earlier versions of Solr will need to be rebuilt.

Factory class: solr.BeiderMorseFilterFactory

Arguments:

nameType: Types of names. Valid values are GENERIC, ASHKENAZI, or SEPHARDIC. If not processing Ashkenazi or Sephardic names, use GENERIC.

ruleType: Types of rules to apply. Valid values are APPROX or EXACT.

concat: Defines if multiple possible matches should be combined with a pipe ("|").

languageSet: The language set to use. The value "auto" will allow the Filter to identify the language, or a comma-separated list can be supplied.

Example:

Classic Filter

This filter takes the output of the Classic Tokenizer and strips periods from acronyms and "'s" from possessives.

Factory class: solr.ClassicFilterFactory

Arguments: None

Example:

In: "I.B.M. cat's can't"

Tokenizer to Filter: "I.B.M", "cat's", "can't"

Out: "IBM", "cat", "can't"

Common Grams Filter

This filter creates word shingles by combining common tokens such as stop words with regular tokens. This is useful for creating phrase queries containing common words, such as "the cat." Solr normally ignores stop words in queried phrases, so searching for "the cat" would return all matches for the word "cat."

Factory class: solr.CommonGramsFilterFactory

Arguments:

words: (a common word file in .txt format) Provide the name of a common word file, such as stopwords.txt.

format: (optional) If the stopwords list has been formatted for Snowball, you can specify format="snowball" so Solr can read the stopwords file.

ignoreCase: (boolean) If true, the filter ignores the case of words when comparing them to the common word file. The default is false.

Example:

In: "the Cat"

Tokenizer to Filter: "the", "Cat"

Out: "the_cat"

Collation Key Filter

Collation allows sorting of text in a language-sensitive way. It is usually used for sorting, but can also be used with advanced searches. We've covered this in much more detail in the section on Unicode Collation.

Daitch-Mokotoff Soundex Filter

Implements the Daitch-Mokotoff Soundex algorithm, which allows identification of similar names, even if they are spelled differently. More information about how this works is available in the section on Phonetic Matching.

Factory class: solr.DaitchMokotoffSoundexFilterFactory

Arguments:

inject : (true/false) If true (the default), then new phonetic tokens are added to the stream. Otherwise, tokens are replaced with the phonetic equivalent. Setting this to false will enable phonetic matching, but the exact spelling of the target word may not match.

Example:

Double Metaphone Filter

This filter creates tokens using the   DoubleMetaphone  encoding algorithm from commons-codec.  For more information, see the Phonetic Matching section.

Factory class: solr.DoubleMetaphoneFilterFactory

Arguments:

inject: (true/false) If true (the default), then new phonetic tokens are added to the stream. Otherwise, tokens are replaced with the phonetic equivalent. Setting this to false will enable phonetic matching, but the exact spelling of the target word may not match.

maxCodeLength: (integer) The maximum length of the code to be generated.

Example:

Default behavior for inject (true): keep the original token and add phonetic token(s) at the same position.

In: "four score and Kuczewski"

Tokenizer to Filter: "four"(1), "score"(2), "and"(3), "Kuczewski"(4)

Out: "four"(1), "FR"(1), "score"(2), "SKR"(2), "and"(3), "ANT"(3), "Kuczewski"(4), "KSSK"(4), "KXFS"(4)

The phonetic tokens have a position increment of 0, which indicates that they are at the same position as the token they were derived from (immediately preceding).  Note that "Kuczewski" has two encodings, which are added at the same position.

Example:

Discard original token (inject="false").

In: "four score and Kuczewski"

Tokenizer to Filter: "four"(1), "score"(2), "and"(3), "Kuczewski"(4)

Out: "FR"(1), "SKR"(2), "ANT"(3), "KSSK"(4), "KXFS"(4)

Note that "Kuczewski" has two encodings, which are added at the same position.

Edge N-Gram Filter

This filter generates edge n-gram tokens of sizes within the given range.

Factory class: solr.EdgeNGramFilterFactory

Arguments:

minGramSize: (integer, default 1) The minimum gram size.

maxGramSize: (integer, default 1) The maximum gram size.

Example:

Default behavior.

In: "four score and twenty"

Tokenizer to Filter: "four", "score", "and", "twenty"

Out: "f", "s", "a", "t"

Example:

A range of 1 to 4.

In: "four score"

Tokenizer to Filter: "four", "score"

Out: "f", "fo", "fou", "four", "s", "sc", "sco", "scor"

Example:

A range of 4 to 6.

In: "four score and twenty"

Tokenizer to Filter: "four", "score", "and", "twenty"

Out: "four", "scor", "score", "twen", "twent", "twenty"

English Minimal Stem Filter

This filter stems plural English words to their singular form.

Factory class: solr.EnglishMinimalStemFilterFactory

Arguments: None

Example:

In: "dogs cats"

Tokenizer to Filter: "dogs", "cats"

Out: "dog", "cat"

Fingerprint Filter

This filter outputs a single token which is a concatenation of the sorted and de-duplicated set of input tokens. This can be useful for clustering/linking use cases.

Factory class: solr.FingerprintFilterFactory

Arguments:

separator: The character used to separate tokens combined into the single output token. Defaults to " " (a space character).

maxOutputTokenSize: The maximum length of the summarized output token. If exceeded, no output token is emitted. Defaults to 1024.

 

Example:

In: "the quick brown fox jumped over the lazy dog"

Tokenizer to Filter: "the", "quick", "brown", "fox", "jumped", "over", "the", "lazy", "dog"

Out: "brown_dog_fox_jumped_lazy_over_quick_the"

Hunspell Stem Filter

The Hunspell Stem Filter provides support for several languages. You must provide the dictionary (.dic) and rules (.aff) files for each language you wish to use with the Hunspell Stem Filter. You can download those language files here. Be aware that your results will vary widely based on the quality of the provided dictionary and rules files. For example, some languages have only a minimal word list with no morphological information. On the other hand, for languages that have no stemmer but do have an extensive dictionary file, the Hunspell stemmer may be a good choice.

Factory class: solr.HunspellStemFilterFactory

Arguments:

dictionary: (required) The path of a dictionary file.

affix: (required) The path of a rules file.

ignoreCase: (boolean) controls whether matching is case sensitive or not. The default is false.

strictAffixParsing: (boolean) controls whether the affix parsing is strict or not. If true, an error while reading an affix rule causes a ParseException, otherwise is ignored. The default is true.

Example:

In: "jump jumping jumped"

Tokenizer to Filter: "jump", "jumping", "jumped"

Out: "jump", "jump", "jump"

Hyphenated Words Filter

This filter reconstructs hyphenated words that have been tokenized as two tokens because of a line break or other intervening whitespace in the field test. If a token ends with a hyphen, it is joined with the following token and the hyphen is discarded. Note that for this filter to work properly, the upstream tokenizer must not remove trailing hyphen characters. This filter is generally only useful at index time.

Factory class: solr.HyphenatedWordsFilterFactory

Arguments: None

Example:

In: "A hyphen- ated word"

Tokenizer to Filter: "A", "hyphen-", "ated", "word"

Out: "A", "hyphenated", "word"

ICU Folding Filter

This filter is a custom Unicode normalization form that applies the foldings specified in Unicode Technical Report 30 in addition to the NFKC_Casefold normalization form as described in ICU Normalizer 2 Filter. This filter is a better substitute for the combined behavior of the ASCII Folding Filter, Lower Case Filter, and ICU Normalizer 2 Filter.

To use this filter, see solr/contrib/analysis-extras/README.txt for instructions on which jars you need to add to your solr_home/lib.

Factory class: solr.ICUFoldingFilterFactory

Arguments: None

Example:

For detailed information on this normalization form, see http://www.unicode.org/reports/tr30/tr30-4.html.

ICU Normalizer 2 Filter

This filter factory normalizes text according to one of five Unicode Normalization Forms as described in Unicode Standard Annex #15:

  • NFC: (name="nfc" mode="compose") Normalization Form C, canonical decomposition
  • NFD: (name="nfc" mode="decompose") Normalization Form D, canonical decomposition, followed by canonical composition
  • NFKC: (name="nfkc" mode="compose") Normalization Form KC, compatibility decomposition
  • NFKD: (name="nfkc" mode="decompose") Normalization Form KD, compatibility decomposition, followed by canonical composition
  • NFKC_Casefold: (name="nfkc_cf" mode="compose") Normalization Form KC, with additional Unicode case folding. Using the ICU Normalizer 2 Filter is a better-performing substitution for the Lower Case Filter and NFKC normalization.

Factory class: solr.ICUNormalizer2FilterFactory

Arguments:

name: (string) The name of the normalization form; nfc, nfd, nfkc, nfkd, nfkc_cf

mode: (string) The mode of Unicode character composition and decomposition; compose or decompose

Example:

For detailed information about these Unicode Normalization Forms, see http://unicode.org/reports/tr15/.

To use this filter, see solr/contrib/analysis-extras/README.txt for instructions on which jars you need to add to your solr_home/lib.

ICU Transform Filter

This filter applies ICU Tranforms to text. This filter supports only ICU System Transforms. Custom rule sets are not supported.

Factory class: solr.ICUTransformFilterFactory

Arguments:

id: (string) The identifier for the ICU System Transform you wish to apply with this filter. For a full list of ICU System Transforms, see http://demo.icu-project.org/icu-bin/translit?TEMPLATE_FILE=data/translit_rule_main.html.

Example:

For detailed information about ICU Transforms, see http://userguide.icu-project.org/transforms/general.

To use this filter, see solr/contrib/analysis-extras/README.txt for instructions on which jars you need to add to your solr_home/lib.

Keep Word Filter

This filter discards all tokens except those that are listed in the given word list. This is the inverse of the Stop Words Filter. This filter can be useful for building specialized indices for a constrained set of terms.

Factory class: solr.KeepWordFilterFactory

Arguments:

words: (required) Path of a text file containing the list of keep words, one per line. Blank lines and lines that begin with "#" are ignored. This may be an absolute path, or a simple filename in the Solr config directory.

ignoreCase: (true/false) If true then comparisons are done case-insensitively. If this argument is true, then the words file is assumed to contain only lowercase words. The default is false.

enablePositionIncrements: if luceneMatchVersion is 4.3 or earlier and enablePositionIncrements="false",  no position holes will be left by this filter when it removes tokens.  This argument is invalid if luceneMatchVersion is 5.0 or later.

Example:

Where keepwords.txt contains:

happy
funny
silly

In: "Happy, sad or funny"

Tokenizer to Filter: "Happy", "sad", "or", "funny"

Out: "funny"

Example:

Same keepwords.txt, case insensitive:

In: "Happy, sad or funny"

Tokenizer to Filter: "Happy", "sad", "or", "funny"

Out: "Happy", "funny"

Example:

Using LowerCaseFilterFactory before filtering for keep words, no ignoreCase flag.

In: "Happy, sad or funny"

Tokenizer to Filter: "Happy", "sad", "or", "funny"

Filter to Filter: "happy", "sad", "or", "funny"

Out: "happy", "funny"

KStem Filter

KStem is an alternative to the Porter Stem Filter for developers looking for a less aggressive stemmer. KStem was written by Bob Krovetz, ported to Lucene by Sergio Guzman-Lara (UMASS Amherst). This stemmer is only appropriate for English language text.

Factory class: solr.KStemFilterFactory

Arguments: None

Example:

In: "jump jumping jumped"

Tokenizer to Filter: "jump", "jumping", "jumped"

Out: "jump", "jump", "jump"

Length Filter

This filter passes tokens whose length falls within the min/max limit specified. All other tokens are discarded.

Factory class: solr.LengthFilterFactory

Arguments:

min: (integer, required) Minimum token length. Tokens shorter than this are discarded.

max: (integer, required, must be >= min) Maximum token length. Tokens longer than this are discarded.

enablePositionIncrements: if luceneMatchVersion is 4.3 or earlier and enablePositionIncrements="false",  no position holes will be left by this filter when it removes tokens.  This argument is invalid if luceneMatchVersion is 5.0 or later.

Example:

In: "turn right at Albuquerque"

Tokenizer to Filter: "turn", "right", "at", "Albuquerque"

Out: "turn", "right"

Lower Case Filter

Converts any uppercase letters in a token to the equivalent lowercase token. All other characters are left unchanged.

Factory class: solr.LowerCaseFilterFactory

Arguments: None

Example:

In: "Down With CamelCase"

Tokenizer to Filter: "Down", "With", "CamelCase"

Out: "down", "with", "camelcase"

Managed Stop Filter

This is specialized version of the  Stop Words Filter Factory that uses a set of stop words that are managed from a REST API.

Arguments:

managed: The name that should be used for this set of stop words in the managed REST API.

Example:

With this configuration the set of words is named "english" and can be managed via /solr/collection_name/schema/analysis/stopwords/english

See Stop Filter for example input/output.

Managed Synonym Filter

This is specialized version of the  Synonym Filter Factory that uses a mapping on synonyms that is managed from a REST API.

Arguments:

managed: The name that should be used for this mapping on synonyms in the managed REST API.

Example:

With this configuration the set of mappings is named "english" and can be managed via /solr/collection_name/schema/analysis/synonyms/english

See Synonym Filter for example input/output.

N-Gram Filter

Generates n-gram tokens of sizes in the given range. Note that tokens are ordered by position and then by gram size.

Factory class: solr.NGramFilterFactory

Arguments:

minGramSize: (integer, default 1) The minimum gram size.

maxGramSize: (integer, default 2) The maximum gram size.

Example:

Default behavior.

In: "four score"

Tokenizer to Filter: "four", "score"

Out: "f", "o", "u", "r", "fo", "ou", "ur", "s", "c", "o", "r", "e", "sc", "co", "or", "re"

Example:

A range of 1 to 4.

In: "four score"

Tokenizer to Filter: "four", "score"

Out: "f", "fo", "fou", "four", "o", "ou", "our", "u", "ur", "r", "s", "sc", "sco", "scor", "c", "co", "cor", "core", "o", "or", "ore", "r", "re", "e"

Example:

A range of 3 to 5.

In: "four score"

Tokenizer to Filter: "four", "score"

Out: "fou", "four", "our", "sco", "scor", "score", "cor", "core", "ore"

Numeric Payload Token Filter

This filter adds a numeric floating point payload value to tokens that match a given type. Refer to the Javadoc for the org.apache.lucene.analysis.Token class for more information about token types and payloads.

Factory class: solr.NumericPayloadTokenFilterFactory

Arguments:

payload: (required) A floating point value that will be added to all matching tokens.

typeMatch: (required) A token type name string. Tokens with a matching type name will have their payload set to the above floating point value.

Example:

In: "bing bang boom"

Tokenizer to Filter: "bing", "bang", "boom"

Out: "bing"[0.75], "bang"[0.75], "boom"[0.75]

Pattern Replace Filter

This filter applies a regular expression to each token and, for those that match, substitutes the given replacement string in place of the matched pattern. Tokens which do not match are passed though unchanged.

Factory class: solr.PatternReplaceFilterFactory

Arguments:

pattern: (required) The regular expression to test against each token, as per java.util.regex.Pattern.

replacement: (required) A string to substitute in place of the matched pattern. This string may contain references to capture groups in the regex pattern. See the Javadoc for java.util.regex.Matcher.

replace: ("all" or "first", default "all") Indicates whether all occurrences of the pattern in the token should be replaced, or only the first.

Example:

Simple string replace:

In: "cat concatenate catycat"

Tokenizer to Filter: "cat", "concatenate", "catycat"

Out: "dog", "condogenate", "dogydog"

Example:

String replacement, first occurrence only:

In: "cat concatenate catycat"

Tokenizer to Filter: "cat", "concatenate", "catycat"

Out: "dog", "condogenate", "dogycat"

Example:

More complex pattern with capture group reference in the replacement. Tokens that start with non-numeric characters and end with digits will have an underscore inserted before the numbers. Otherwise the token is passed through.

In: "cat foo1234 9987 blah1234foo"

Tokenizer to Filter: "cat", "foo1234", "9987", "blah1234foo"

Out: "cat", "foo_1234", "9987", "blah1234foo"

Phonetic Filter

This filter creates tokens using one of the phonetic encoding algorithms in the org.apache.commons.codec.language package.  For more information, see the section on Phonetic Matching.

Factory class: solr.PhoneticFilterFactory

Arguments:

encoder: (required) The name of the encoder to use. The encoder name must be one of the following (case insensitive): "DoubleMetaphone", "Metaphone", "Soundex", "RefinedSoundex", "Caverphone" (v2.0), "ColognePhonetic", or "Nysiis".

inject: (true/false) If true (the default), then new phonetic tokens are added to the stream. Otherwise, tokens are replaced with the phonetic equivalent. Setting this to false will enable phonetic matching, but the exact spelling of the target word may not match.

maxCodeLength: (integer) The maximum length of the code to be generated by the Metaphone or Double Metaphone encoders.

Example:

Default behavior for DoubleMetaphone encoding.

In: "four score and twenty"

Tokenizer to Filter: "four"(1), "score"(2), "and"(3), "twenty"(4)

Out: "four"(1), "FR"(1), "score"(2), "SKR"(2), "and"(3), "ANT"(3), "twenty"(4), "TNT"(4)

The phonetic tokens have a position increment of 0, which indicates that they are at the same position as the token they were derived from (immediately preceding).

Example:

Discard original token.

In: "four score and twenty"

Tokenizer to Filter: "four"(1), "score"(2), "and"(3), "twenty"(4)

Out: "FR"(1), "SKR"(2), "ANT"(3), "TWNT"(4)

Example:

Default Soundex encoder.

In: "four score and twenty"

Tokenizer to Filter: "four"(1), "score"(2), "and"(3), "twenty"(4)

Out: "four"(1), "F600"(1), "score"(2), "S600"(2), "and"(3), "A530"(3), "twenty"(4), "T530"(4)

Porter Stem Filter

This filter applies the Porter Stemming Algorithm for English. The results are similar to using the Snowball Porter Stemmer with the language="English" argument. But this stemmer is coded directly in Java and is not based on Snowball. It does not accept a list of protected words and is only appropriate for English language text. However, it has been benchmarked as four times faster than the English Snowball stemmer, so can provide a performance enhancement.

Factory class: solr.PorterStemFilterFactory

Arguments: None

Example:

In: "jump jumping jumped"

Tokenizer to Filter: "jump", "jumping", "jumped"

Out: "jump", "jump", "jump"

Remove Duplicates Token Filter

The filter removes duplicate tokens in the stream. Tokens are considered to be duplicates if they have the same text and position values.

Factory class: solr.RemoveDuplicatesTokenFilterFactory

Arguments: None

Example:

One example of where RemoveDuplicatesTokenFilterFactory is in situations where a synonym file is being used in conjuntion with a stemmer causes some synonyms to be reduced to the same stem. Consider the following entry from a synonyms.txt file:

When used in the following configuration:

In: "Watch TV"

Tokenizer to Synonym Filter: "Watch"(1) "TV"(2)

Synonym Filter to Stem Filter: "Watch"(1) "Television"(2) "Televisions"(2) "TV"(2) "TVs"(2)

Stem Filter to Remove Dups Filter: "Watch"(1) "Television"(2) "Television"(2) "TV"(2) "TV"(2)

Out: "Watch"(1) "Television"(2) "TV"(2)

Reversed Wildcard Filter

This filter reverses tokens to provide faster leading wildcard and prefix queries. Tokens without wildcards are not reversed.

Factory class: solr.ReversedWildcardFilterFactory

Arguments:

withOriginal (boolean) If true, the filter produces both original and reversed tokens at the same positions. If false, produces only reversed tokens.

maxPosAsterisk (integer, default = 2) The maximum position of the asterisk wildcard ('*') that triggers the reversal of the query term. Terms with asterisks at positions above this value are not reversed.

maxPosQuestion (integer, default = 1) The maximum position of the question mark wildcard ('?') that triggers the reversal of query term. To reverse only pure suffix queries (queries with a single leading asterisk), set this to 0 and maxPosAsterisk to 1.

maxFractionAsterisk (float, default = 0.0) An additional parameter that triggers the reversal if asterisk ('*') position is less than this fraction of the query token length.

minTrailing (integer, default = 2) The minimum number of trailing characters in a query token after the last wildcard character. For good performance this should be set to a value larger than 1.

Example:

In: "*foo *bar"

Tokenizer to Filter: "*foo", "*bar"

Out: "oof*", "rab*"

Shingle Filter

This filter constructs shingles, which are token n-grams, from the token stream. It combines runs of tokens into a single token.

Factory class: solr.ShingleFilterFactory

Arguments:

minShingleSize: (integer, default 2) The minimum number of tokens per shingle.

maxShingleSize: (integer, must be >= 2, default 2) The maximum number of tokens per shingle.

outputUnigrams: (true/false) If true (the default), then each individual token is also included at its original position.

outputUnigramsIfNoShingles: (true/false) If false (the default), then individual tokens will be output if no shingles are possible.

tokenSeparator: (string, default is " ") The default string to use when joining adjacent tokens to form a shingle.

Example:

Default behavior.

In: "To be, or what?"

Tokenizer to Filter: "To"(1), "be"(2), "or"(3), "what"(4)

Out: "To"(1), "To be"(1), "be"(2), "be or"(2), "or"(3), "or what"(3), "what"(4)

Example:

A shingle size of four, do not include original token.

In: "To be, or not to be."

Tokenizer to Filter: "To"(1), "be"(2), "or"(3), "not"(4), "to"(5), "be"(6)

Out: "To be"(1), "To be or"(1), "To be or not"(1), "be or"(2), "be or not"(2), "be or not to"(2), "or not"(3), "or not to"(3), "or not to be"(3), "not to"(4), "not to be"(4), "to be"(5)

Snowball Porter Stemmer Filter

This filter factory instantiates a language-specific stemmer generated by Snowball. Snowball is a software package that generates pattern-based word stemmers. This type of stemmer is not as accurate as a table-based stemmer, but is faster and less complex. Table-driven stemmers are labor intensive to create and maintain and so are typically commercial products.

Solr contains Snowball stemmers for Armenian, Basque, Catalan, Danish, Dutch, English, Finnish, French, German, Hungarian, Italian, Norwegian, Portuguese, Romanian, Russian, Spanish, Swedish and Turkish. For more information on Snowball, visit http://snowball.tartarus.org/.

StopFilterFactory, CommonGramsFilterFactory, and CommonGramsQueryFilterFactory can optionally read stopwords in Snowball format (specify format="snowball" in the configuration of those FilterFactories).

Factory class: solr.SnowballPorterFilterFactory

Arguments:

language: (default "English") The name of a language, used to select the appropriate Porter stemmer to use. Case is significant. This string is used to select a package name in the "org.tartarus.snowball.ext" class hierarchy.

protected: Path of a text file containing a list of protected words, one per line. Protected words will not be stemmed. Blank lines and lines that begin with "#" are ignored. This may be an absolute path, or a simple file name in the Solr config directory.

Example:

Default behavior:

In: "flip flipped flipping"

Tokenizer to Filter: "flip", "flipped", "flipping"

Out: "flip", "flip", "flip"

Example:

French stemmer, English words:

In: "flip flipped flipping"

Tokenizer to Filter: "flip", "flipped", "flipping"

Out: "flip", "flipped", "flipping"

Example:

Spanish stemmer, Spanish words:

In: "cante canta"

Tokenizer to Filter: "cante", "canta"

Out: "cant", "cant"

Standard Filter

This filter removes dots from acronyms and the substring "'s" from the end of tokens. This filter depends on the tokens being tagged with the appropriate term-type to recognize acronyms and words with apostrophes.

Factory class: solr.StandardFilterFactory

Arguments: None

This filter is no longer operational in Solr when the luceneMatchVersion (in solrconfig.xml) is higher than "3.1".

Stop Filter

This filter discards, or stops analysis of, tokens that are on the given stop words list. A standard stop words list is included in the Solr config directory, named stopwords.txt, which is appropriate for typical English language text.

Factory class: solr.StopFilterFactory

Arguments:

words: (optional) The path to a file that contains a list of stop words, one per line. Blank lines and lines that begin with "#" are ignored. This may be an absolute path, or path relative to the Solr config directory.

format: (optional) If the stopwords list has been formatted for Snowball, you can specify format="snowball" so Solr can read the stopwords file.

ignoreCase: (true/false, default false) Ignore case when testing for stop words. If true, the stop list should contain lowercase words.

enablePositionIncrements: if luceneMatchVersion is 4.4 or earlier and enablePositionIncrements="false",  no position holes will be left by this filter when it removes tokens.  This argument is invalid if luceneMatchVersion is 5.0 or later.

Example:

Case-sensitive matching, capitalized words not stopped. Token positions skip stopped words.

In: "To be or what?"

Tokenizer to Filter: "To"(1), "be"(2), "or"(3), "what"(4)

Out: "To"(1), "what"(4)

Example:

In: "To be or what?"

Tokenizer to Filter: "To"(1), "be"(2), "or"(3), "what"(4)

Out: "what"(4)

Suggest Stop Filter

Like Stop Filter, this filter discards, or stops analysis of, tokens that are on the given stop words list. Suggest Stop Filter differs from Stop Filter in that it will not remove the last token unless it is followed by a token separator. For example, a query "find the" would preserve the 'the' since it was not followed by a space, punctuation etc., and mark it as a KEYWORD so that following filters  will not change or remove it.  By contrast, a query like "find the popsicle" would remove "the" as a stopword, since it's followed by a space.  When using one of the analyzing suggesters, you would normally use the ordinary StopFilterFactory in your index analyzer and then SuggestStopFilter in your query analyzer.

Factory class: solr.SuggestStopFilterFactory

Arguments:

words: (optional; default: StopAnalyzer#ENGLISH_STOP_WORDS_SET )  The name of a stopwords file to parse.

format: (optional; default: wordset) Defines how the words file will be parsed. If words is not specified, then format must not be specified. The valid values for the format option are:

  • wordset: This is the default format, which supports one word per line (including any intra-word whitespace) and allows whole line comments begining with the "#" character. Blank lines are ignored.
  • snowball: This format allows for multiple words specified on each line, and trailing comments may be specified using the vertical line ("|"). Blank lines are ignored.

ignoreCase: (optional; default: false) If true, matching is case-insensitive.

Example:

 

In: "The The"

Tokenizer to Filter: "the"(1), "the"(2)

Out: "the"(2)

Synonym Filter

This filter does synonym mapping. Each token is looked up in the list of synonyms and if a match is found, then the synonym is emitted in place of the token. The position value of the new tokens are set such they all occur at the same position as the original token.

Factory class: solr.SynonymFilterFactory

Arguments:

synonyms: (required) The path of a file that contains a list of synonyms, one per line. In the (default) solr format - see the format argument below for alternatives - blank lines and lines that begin with "#" are ignored. This may be an absolute path, or path relative to the Solr config directory.  There are two ways to specify synonym mappings:

  • A comma-separated list of words. If the token matches any of the words, then all the words in the list are substituted, which will include the original token.
  • Two comma-separated lists of words with the symbol "=>" between them. If the token matches any word on the left, then the list on the right is substituted. The original token will not be included unless it is also in the list on the right.

ignoreCase: (optional; default: false) If true, synonyms will be matched case-insensitively.

expand: (optional; default: true) If true, a synonym will be expanded to all equivalent synonyms. If false, all equivalent synonyms will be reduced to the first in the list.

format: (optional; default: solr) Controls how the synonyms will be parsed. The short names solr (for SolrSynonymParser) and wordnet (for WordnetSynonymParser ) are supported, or you may alternatively supply the name of your own  SynonymMap.Builder  subclass.

tokenizerFactory: (optional; default: WhitespaceTokenizerFactory) The name of the tokenizer factory to use when parsing the synonyms file.  Arguments with the name prefix "tokenizerFactory." will be supplied as init params to the specified tokenizer factory.  Any arguments not consumed by the synonym filter factory, including those without the "tokenizerFactory." prefix, will also be supplied as init params to the tokenizer factory. If tokenizerFactory is specified, then analyzer may not be, and vice versa.

analyzer: (optional; default: WhitespaceTokenizerFactory) The name of the analyzer class to use when parsing the synonyms file.  If analyzer is specified, then tokenizerFactory may not be, and vice versa.

For the following examples, assume a synonyms file named mysynonyms.txt:

Example:

In: "teh small couch"

Tokenizer to Filter: "teh"(1), "small"(2), "couch"(3)

Out: "the"(1), "tiny"(2), "teeny"(2), "weeny"(2), "couch"(3), "sofa"(3), "divan"(3)

Example:

In: "teh ginormous, humungous sofa"

Tokenizer to Filter: "teh"(1), "ginormous"(2), "humungous"(3), "sofa"(4)

Out: "the"(1), "large"(2), "large"(3), "couch"(4), "sofa"(4), "divan"(4)

Token Offset Payload Filter

This filter adds the numeric character offsets of the token as a payload value for that token.

Factory class: solr.TokenOffsetPayloadTokenFilterFactory

Arguments: None

Example:

In: "bing bang boom"

Tokenizer to Filter: "bing", "bang", "boom"

Out: "bing"[0,4], "bang"[5,9], "boom"[10,14]

Trim Filter

This filter trims leading and/or trailing whitespace from tokens. Most tokenizers break tokens at whitespace, so this filter is most often used for special situations.

Factory class: solr.TrimFilterFactory

Arguments:

updateOffsets: if luceneMatchVersion is 4.3 or earlier and updateOffsets="true",  trimmed tokens' start and end offsets will be updated to those of the first and last characters (plus one) remaining in the token.  This argument is invalid if luceneMatchVersion is 5.0 or later.

Example:

The PatternTokenizerFactory configuration used here splits the input on simple commas, it does not remove whitespace.

In: "one, two , three ,four "

Tokenizer to Filter: "one", " two ", " three ", "four "

Out: "one", "two", "three", "four"

Type As Payload Filter

This filter adds the token's type, as an encoded byte sequence, as its payload.

Factory class: solr.TypeAsPayloadTokenFilterFactory

Arguments: None

Example:

In: "Pay Bob's I.O.U."

Tokenizer to Filter: "Pay", "Bob's", "I.O.U."

Out: "Pay"[<ALPHANUM>], "Bob's"[<APOSTROPHE>], "I.O.U."[<ACRONYM>]

Type Token Filter

This filter blacklists or whitelists a specified list of token types, assuming the tokens have type metadata associated with them. For example, the UAX29 URL Email Tokenizer emits "<URL>" and "<EMAIL>" typed tokens, as well as other types. This filter would allow you to pull out only e-mail addresses from text as tokens, if you wish.

Factory class: solr.TypeTokenFilterFactory

Arguments:

types: Defines the location of a file of types to filter.

useWhitelist: If true, the file defined in types should be used as include list. If false, or undefined, the file defined in types is used as a blacklist.

enablePositionIncrements: if luceneMatchVersion is 4.3 or earlier and enablePositionIncrements="false",  no position holes will be left by this filter when it removes tokens.  This argument is invalid if luceneMatchVersion is 5.0 or later.

Example:

Word Delimiter Filter

This filter splits tokens at word delimiters. The rules for determining delimiters are determined as follows:

  • A change in case within a word: "CamelCase" -> "Camel", "Case". This can be disabled by setting splitOnCaseChange="0".
  • A transition from alpha to numeric characters or vice versa: "Gonzo5000" -> "Gonzo", "5000" "4500XL" -> "4500", "XL". This can be disabled by setting splitOnNumerics="0".
  • Non-alphanumeric characters (discarded): "hot-spot" -> "hot", "spot"
  • A trailing "'s" is removed: "O'Reilly's" -> "O", "Reilly"
  • Any leading or trailing delimiters are discarded: "--hot-spot--" -> "hot", "spot"

Factory class: solr.WordDelimiterFilterFactory

Arguments:

generateWordParts: (integer, default 1) If non-zero, splits words at delimiters. For example:"CamelCase", "hot-spot" -> "Camel", "Case", "hot", "spot"

generateNumberParts: (integer, default 1) If non-zero, splits numeric strings at delimiters:"1947-32" ->"1947", "32"

splitOnCaseChange: (integer, default 1) If 0, words are not split on camel-case changes:"BugBlaster-XL" -> "BugBlaster", "XL". Example 1 below illustrates the default (non-zero) splitting behavior.

splitOnNumerics: (integer, default 1) If 0, don't split words on transitions from alpha to numeric:"FemBot3000" -> "Fem", "Bot3000"

catenateWords: (integer, default 0) If non-zero, maximal runs of word parts will be joined: "hot-spot-sensor's" -> "hotspotsensor"

catenateNumbers: (integer, default 0) If non-zero, maximal runs of number parts will be joined: 1947-32" -> "194732"

catenateAll: (0/1, default 0) If non-zero, runs of word and number parts will be joined: "Zap-Master-9000" -> "ZapMaster9000"

preserveOriginal: (integer, default 0) If non-zero, the original token is preserved: "Zap-Master-9000" -> "Zap-Master-9000", "Zap", "Master", "9000"

protected: (optional) The pathname of a file that contains a list of protected words that should be passed through without splitting.

stemEnglishPossessive: (integer, default 1) If 1, strips the possessive "'s" from each subword.

Example:

Default behavior. The whitespace tokenizer is used here to preserve non-alphanumeric characters.

In: "hot-spot RoboBlaster/9000 100XL"

Tokenizer to Filter: "hot-spot", "RoboBlaster/9000", "100XL"

Out: "hot", "spot", "Robo", "Blaster", "9000", "100", "XL"

Example:

Do not split on case changes, and do not generate number parts. Note that by not generating number parts, tokens containing only numeric parts are ultimately discarded.

In: "hot-spot RoboBlaster/9000 100-42"

Tokenizer to Filter: "hot-spot", "RoboBlaster/9000", "100-42"

Out: "hot", "spot", "RoboBlaster", "9000"

Example:

Concatenate word parts and number parts, but not word and number parts that occur in the same token.

In: "hot-spot 100+42 XL40"

Tokenizer to Filter: "hot-spot"(1), "100+42"(2), "XL40"(3)

Out: "hot"(1), "spot"(2), "hotspot"(2), "100"(3), "42"(4), "10042"(4), "XL"(5), "40"(6)

Example:

Concatenate all. Word and/or number parts are joined together.

In: "XL-4000/ES"

Tokenizer to Filter: "XL-4000/ES"(1)

Out: "XL"(1), "4000"(2), "ES"(3), "XL4000ES"(3)

Example:

Using a protected words list that contains "AstroBlaster" and "XL-5000" (among others).

In: "FooBar AstroBlaster XL-5000 ==ES-34-"

Tokenizer to Filter: "FooBar", "AstroBlaster", "XL-5000", "==ES-34-"

Out: "FooBar", "FooBar", "AstroBlaster", "XL-5000", "ES", "34"

 

  • No labels

17 Comments

  1. Small typo in four places : solr.PatternReplaceFilter => solr.PatternReplaceFilterFactory

  2. Edge N-Gram Filter ... A range of 4 to 6 ... Out: "four", "sco", "scor", "twen", "twent", "twenty"

    Sould be : ... "scor", "score" ...

    1. Thanks Laurent, I've fixed that example.

  3. We should add description of  preserveOriginal parameter to solr.ASCIIFoldingFilterFactory.

  4. In WordDelimiterFilter generateWordParts the example "CamelCase" is conflicting with splitOnCaseChange settings. Please make it clear on  generateWordParts description as this will generate words based on default delimiters and other options set below or something like that...

  5. I think this should be:

    Unlike tokenizers, filters only ...

    The class attribute names a factory class that will instantiate a filter object as needed. Filter factory classes must implement the org.apache.solr.analysis.TokenFilterFactoryinterface. Like tokenizers, filters are also instances of TokenStream and thus are producers of tokens. Unlike tokenizers, filters also consume tokens from a TokenStream. This allows you to mix and match filters, in any order you prefer, downstream of a tokenizer.

    1. but filter's don't "only" consume tokens, they "also" consume tokens in addition to producing tokens (which is what the previous sentence says)

       

      (there's probably a better way to word that entire paragraph, but changing "also" to "only" isn't accurate)

       

       

  6. Cwiki does not mention that it is possible set luceneMatchVersion parameter when defining filters. We can give this example in word delimiter for example.

  7. Hmmm...

     

    EnglishPossessiveFilterFactory isn't mentioned here.  probably some others, need to do a full audit.

  8. There is small a typo on ICU Normalizer 2 Filter example:

    <filter class="solr.ICUNormalizer2FilterFactory" name="nkc_cf" mode="compose"/>  <-- invalid name, it should be nfkc_cf instead.  It will cause some errors initializing.

    1. Fixed.  Thanks for letting us know about the problem.

  9. The second example for NGramFilter has been copied verbatim from the example for EdgeNgram, but the behavior isn't the same:

        <filter class="solr.NGramFilterFactory" minGramSize="1" maxGramSize="4"/>

    Should give "f", "fo", "fou", "four", "o", "ou", "our", "u", "ur", "r", "s", "sc", "sco", "scor", "c", "co", "cor", "core", "o", "or", "ore", "r", "re", "e" .. if I managed to type it correctly.

    1. This is not the correct place to obtain support for Solr.  We have a mailing list and an IRC channel.  Please use one of those resources.  These comments are for discussions about the documentation itself.

      http://lucene.apache.org/solr/community.html

      1. This isn't a support request. The example used on "NGramFilter" has been copied verbatim from "EdgeNgram", without updating the output produced by the NGramFilter compared to what the EdgeNGramFilter produces. As far as I can tell the current second example under "NGramFilter" is wrong (it has the output produced by the EdgeNGramFilter, while the other two are correct).

        1. My apologies.  I thought you were saying that you were getting a different result from the NGram filter in practice than you expected to get.  It looks to me like your list of terms is correct.  I will adjust the documentation.