Package org.apache.lucene.analysis.pt
Class RSLPStemmerBase
- java.lang.Object
-
- org.apache.lucene.analysis.pt.RSLPStemmerBase
-
- Direct Known Subclasses:
GalicianMinimalStemmer,GalicianStemmer,PortugueseMinimalStemmer,PortugueseStemmer
public abstract class RSLPStemmerBase extends java.lang.ObjectBase class for stemmers that use a set of RSLP-like stemming steps.RSLP (Removedor de Sufixos da Lingua Portuguesa) is an algorithm designed originally for stemming the Portuguese language, described in the paper A Stemming Algorithm for the Portuguese Language, Orengo et. al.
Since this time a plural-only modification (RSLP-S) as well as a modification for the Galician language have been implemented. This class parses a configuration file that describes
RSLPStemmerBase.Steps, where each Step contains a set ofRSLPStemmerBase.Rules.The general rule format is:
{ "suffix", N, "replacement", { "exception1", "exception2", ...}}
where:suffixis the suffix to be removed (such as "inho").Nis the min stem size, where stem is defined as the candidate stem after removing the suffix (but before appending the replacement!)replacementis an optimal string to append after removing the suffix. This can be the empty string.exceptionsis an optional list of exceptions, patterns that should not be stemmed. These patterns can be specified as whole word or suffix (ends-with) patterns, depending upon the exceptions format flag in the step header.
A step is an ordered list of rules, with a structure in this format:
{ "name", N, B, { "cond1", "cond2", ... } ... rules ... };
where:nameis a name for the step (such as "Plural").Nis the min word size. Words that are less than this length bypass the step completely, as an optimization. Note: N can be zero, in this case this implementation will automatically calculate the appropriate value from the underlying rules.Bis a "boolean" flag specifying how exceptions in the rules are matched. A value of 1 indicates whole-word pattern matching, a value of 0 indicates that exceptions are actually suffixes and should be matched with ends-with.condsare an optional list of conditions to enter the step at all. If the list is non-empty, then a word must end with one of these conditions or it will bypass the step completely as an optimization.
- See Also:
- RSLP description
-
-
Constructor Summary
Constructors Constructor Description RSLPStemmerBase()
-