how to use any statistical stemmer like YASS or GRAS in place of Porter stemmer in TERRIER Information Retreival system

Question

Terrier IR system uses Porter Stemmer by default. How can we use statistical stemmers output in Terrier? i have generated the stem list using statistical stemmer and want to embed it in TERRIER IR.

Answer 1

You have to create a class extending StemmerTermPipeline into the org.terrier.terms package.

For instance:

public class StatisticalStemmer extends StemmerTermPipeline {

public StatisticalStemmer(TermPipeline next) {
    super(next);
}

@Override
public String stem(String word) {
   // your method implementation
}

}

Afterwards, you need to recompile the core component and to substitute the terrier-4.0-core.jar file in the lib directory.

Lastly, you need to update the Term Pipeline in the property file:

termpipelines=Stopwords,StatisticalStemmer

In this way, Terrier will use you stemmer in place of the PorterStemmer.