Boost based on indexed field

Question

I have some fields that are ordered in the index (e.g. lexographical Strings). I'd like to give a higher boost/score to the higher values.

What is the preferred way to do this in Lucene 4? (obtaining the exact value through a hit to the IndexCache is suboptimal)

(There are plently of Google hits for earlier versions of Lucene, but the API has changed substantially with great improvements to the indexing system.)

I assume simply sorting on the field's value would not be adequate? Also, is it acceptable to use field-level (that is, index time) boosting for this? — femtoRgon
I'm currently using index time boosting (but, for various reasons, this only works because we re-index regularly... e.g. to boost "recent" documents). I don't know what you mean by "sorting on the field's value". If there were a way to use the natural index order of the fields, that would be perfect: avoiding actually looking at the document is critical (as with any efficient Lucene query). — fommil
I've noted a dramatic slow-down in insertion speed when doing the index boosting. In Lucene 4, index-time boosting means looping through all the indexable fields and manually boosting them. — fommil

femtoRgon femtoRgon · Accepted Answer · 2013-07-18T18:00:47

You can sort on a fields value lexicographically by passing a Sort into your call to IndexSearcher.search.

SortField primarySort = new SortField("field", SortField.Type.STRING);
Sort sort = new Sort(primarySort, SortField.FIELD_SCORE);
searcher.search(query, hits, sort);

That will sort, first, on the lexicographic ordering of the given field, then by relevance score. You can add as many sort fields as you like when constructing your Sort.

Boost based on indexed field

1 Answers