IR-part2

and average them evaluates performance at all recall

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Google was using over 200 such features. Introduc)on to Informa)on Retrieval How to combine features to assign a relevance score to a document? §༊  Given lots of relevant features… §༊  You can con*nue to hand- engineer retrieval scores §༊  Or, you can build a classifier to learn weights for the features §༊  Requires: labeled training data §༊  This is the “learning to rank” approach, which has become a hot area in recent years §༊  I only provide an elementary introduc*on here Introduc)on to Informa)on Retrieval Sec. 15.4.1 Simple example: Using classifica*on for ad hoc IR §༊  Collect a training corpus of (q, d, r) triples §༊  Relevance r is here binary (but may be mul*class, with 3–7 values) §༊  Document is represented by a feature vector §༊  x = (α, ω) α is cosine similarity, ω is minimum query window size §༊  ω is the the shortest text span that includes all query words §༊  Query term proximity is a very important...
View Full Document

This document was uploaded on 02/14/2014.

Ask a homework question - tutors are online