lecture7-vectorspace-handout-6-per

Lecture7-vectorspace-handout-6-per

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: + cosine(q,d)   Denote this by g(d)   Thus, a quan*ty like the number of cita*ons is scaled into [0,1]   Exercise: suggest a formula for this.   Can use some other linear combina*on   Indeed, any func*on of the two signals of user happiness – more later   Now we seek the top K docs by net score 4 Introduc)on to Informa)on Retrieval Sec. 7.1.4 Introduc)on to Informa)on Retrieval Sec. 7.1.4 Top K by net score – fast methods Why order pos*ngs by g(d)?   First idea: Order all pos*ngs by g(d)   Key: this is a common ordering for all pos*ngs   Thus, can concurrently traverse query terms pos*ngs for   Under g(d) ­ordering, top ­scoring docs likely to appear early in pos*ngs traversal   In *me ­bound applica*ons (say, we have to return whatever search results we can in 50 ms), this allows us to stop pos*ngs traversal early   Pos*ngs intersec*on   Cosine score computa*on   Short of compu*ng scores for all docs in pos*ngs   Exercise: write pseudocode for cosine score computa*on if pos*ngs are ordered by g(d) Introduc)on to Informa)on Retrieval Sec. 7.1.4 Introduc)on to Informa)on Retrieval Sec. 7.1.4 Champion lists in g(d) ­ordering High and low lists   Can combine champion lists with g(d) ­ordering   Maintain for each term a champion list of the r docs with highest g(d) + I ­idftd   Seek top ­K results from only the docs in these champion lists   For each term, we maintain two pos*ngs lists called high and low   Think of high as the champion list   When traversing pos*ngs on a query, only traverse high lists first   If we get more than K docs, select the top K and stop   Else proceed to get docs from the low lists   Can be used even for simple cosine scores, without global quality g(d)  ...
View Full Document

This document was uploaded on 02/26/2014.

Ask a homework question - tutors are online