PageRank-without-hyperlinks

PageRank-without-hyperlinks - ± Weights on edges ±...

Info iconThis preview shows pages 1–13. Sign up to view the full content.

View Full Document Right Arrow Icon
PageRank without hyperlinks: Structural re-ranking using links induced by language models Oren Kurland and Lillian Lee Presentation by Yang Yu
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Related Work ± Query-dependent clustering ± Do not directly induces an obvious ranking of doc. ± Techniques based on graphs ± Similarly used in text summarization
Background image of page 2
Main Idea ± Using language model to improve IR performance ± Generation links , which are based on the probability assigned by the language model induced from one document ± Structural re-ranking ± The analogy between hyperlinks and generation links is not perfect. ± Compute an ordering not of the entire corpus but of a set.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Structural Re-ranking ± Generation Graphs ± pd ( · ) denotes the smoothed unigram language model induced from d ± Example d 1: Toronto Sheffield Salvador d 2: Salvador Salvador Salvador ± Definition 1: The top α generators of a document d D init ± Definition 2: The offspring of a document d D init {o D init : d TopGen ( o ) }
Background image of page 4
Structural Re-ranking ± Generation Graphs (cont.)
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 6
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 8
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 10
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Background image of page 12
Background image of page 13
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ± Weights on edges ± Definition 3: • Weights on smoothed graph Structural Re-ranking ± Computing Graph Centrality ± Uniform Influx ± Weighted Influx ± Recursive Uniform Influx ± Recursive Weighted Influx Structural Re-ranking ± Incorporating initial scores ± Uniform Influx + LM ± Weighted Influx + LM , ± Recursive Uniform Influx + LM ± Recursive Weighted Influx + LM . Structural Re-ranking ± Generation Probabilities: ± Maximum-likelihood estimate (MLE) of w with respect to x ± Dirichlet-smoothed version : Structural Re-ranking ± Generation Probabilities (cont.) ± Kullback-Leibler divergence D Evaluation 8 8 0 8 8 8 4 6 3 7 4 64/96 7 5 10 7 29/48 U-In+LM > U-In = 10 W-In+Lm > W-In = 8 R-U-In+LM > R-U-In = 11 R-W-In+LM > R-W-In =10 39/48 Evaluation ± Links based on the vector-space model Evaluation ± Non-structural re-ranking Questions and Comments...
View Full Document

This note was uploaded on 08/06/2008 for the course CSE 450 taught by Professor Davison during the Spring '08 term at Lehigh University .

Page1 / 13

PageRank-without-hyperlinks - ± Weights on edges ±...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online