Term Vector Frequency in Lucene 4.0

Question

I'm upgrading from Lucene 3.6 to Lucene 4.0-beta. In Lucene 3.x, the IndexReader contains a method IndexReader.getTermFreqVectors() , which I can use to extract the frequency of each term in a given document and field.

This method is now replaced by IndexReader.getTermVectors() , which returns Terms . How can I make use of this (or probably other methods) to extract the term frequency in a document and a field?

Answer 1

Perhaps this will help you:

// get terms vectors for one document and one field
Terms terms = reader.getTermVector(docID, "fieldName"); 

if (terms != null && terms.size() > 0) {
    // access the terms for this field
    TermsEnum termsEnum = terms.iterator(null); 
    BytesRef term = null;

    // explore the terms for this field
    while ((term = termsEnum.next()) != null) {
        // enumerate through documents, in this case only one
        DocsEnum docsEnum = termsEnum.docs(null, null); 
        int docIdEnum;
        while ((docIdEnum = docsEnum.nextDoc()) != DocIdSetIterator.NO_MORE_DOCS) {
            // get the term frequency in the document 
            System.out.println(term.utf8ToString()+ " " + docIdEnum + " " + docsEnum.freq()); 
        }
    }
}

Answer 2

See this related question , specificially

Terms vector = reader.getTermVector(docId, CONTENT);
TermsEnum termsEnum = null;
termsEnum = vector.iterator(termsEnum);
Map<String, Integer> frequencies = new HashMap<>();
BytesRef text = null;
while ((text = termsEnum.next()) != null) {
    String term = text.utf8ToString();
    int freq = (int) termsEnum.totalTermFreq();
    frequencies.put(term, freq);
    terms.add(term);
}

Answer 3

There is various documentation on how to use the flexible indexing apis:

• http://lucene.apache.org/core/4_0_0-BETA/MIGRATE.html
• https://builds.apache.org/job/Lucene-Artifacts-4.x/javadoc/core/org/apache/lucene/index/package-summary.html#postings

Accessing the Fields/Terms for a documents term vectors is the exact same API you use for accessing the postings lists, since term vectors are really just a miniature inverted index for just that one document.

So its perfectly OK to use all those examples as-is, though you can make some shortcuts since you know there is only ever one document in this "miniature inverted index". eg if you just want to get the frequency of a term you can just seek to it and use the aggregate statistics like totalTermFreq (see https://builds.apache.org/job/Lucene-Artifacts-4.x/javadoc/core/org/apache/lucene/index/package-summary.html#stats ), rather than actually opening a DocsEnum that will only enumerate over a single document.

Answer 4

I have this working on my Lucene 4.2 index. This is a small test program that works for me.

try {
    directory[0] = new SimpleFSDirectory(new File(test1));
    directory[1] = new SimpleFSDirectory(new File(test2));
    directory[2] = new SimpleFSDirectory(new File(test3));
    directoryReader[0] = DirectoryReader.open(directory[0]);
    directoryReader[1] = DirectoryReader.open(directory[1]);
    directoryReader[2] = DirectoryReader.open(directory[2]);

    if (!directoryReader[2].isCurrent()) {
        directoryReader[2] = DirectoryReader.openIfChanged(directoryReader[2]);
    }
    MultiReader mr = new MultiReader(directoryReader);

    TermStats[] stats=null;
    try {
        stats = HighFreqTerms.getHighFreqTerms(mr, 100, "My Term");
    } catch (Exception e1) {
        e1.printStackTrace();
        return;
    }

    for (TermStats termstat : stats) {
        System.out.println("IBI_body: " + termstat.termtext.utf8ToString() +
            ", docFrequency: " + termstat.docFreq);
    }
}