July 2017
Beginner to intermediate
715 pages
17h 3m
English
In this chapter, we already mentioned that Lucene can be customized, and we already took a look at how to do preprocessing outside of Lucene and then seamlessly integrate the results in the Lucene workflow.
When it comes to reranking search results, the situation is more or less similar. The common approach to this is taking the Lucene ranking as is and retrieve the top 100 (or more) results. Then we take these already retrieved documents and apply the ranking model to this for reordering.
If we have such a reranking model, we need to make sure that we store all the data we used for training. In our case, it was a QueryDocumentPair class from which we extracted the relevance features. So let's create an index:
FSDirectory ...
Read now
Unlock full access