Using ODP Metadata to Personalize Search

1
Using ODP Metadata to Personalize Search

• Presented by Lan Nie
• 09/21/2005, Lehigh University

2
Introduction

• ODP metadata
• 4 million sites, 590,000 categories
• Tree Structure
• Categories inner node
• Pages leaf node, high quality, representative
• Using ODP Metadata to personalize Search
• 4 billion vs. 4 million
• Using ODP Metadata for personalized search
• Is biasing possible in the ODP context?
• Extend ODP classifications from its current 4
  million to a 4 billion Web automatically by
  biasing

3
Using ODP Metadata For Personalized Search

• User Profile several topics from ODP selected by
  user
• Personalized Search
• Send Q to a search Engine S(E.g., Google, ODP
  Search)
• ResURLs returned by S
• For i 1 to size(Res)
• DistiDistance(Resi, Prof)
• Resort Res based on Dist
• Representation
• Both user profile and URL(50 in Google
  directory) can be represented as a set of nodes
  in the directory tree
• Distance ( Profile, URL)
• Minimum distance between the 2 set of nodes.

4

• Naïve Distances
• Minimum tree distance
• Intra-topic links
• Subsumer
• Graph shortest path
• Inter-topic links

• Complex Distance
• The bigger the subsumers depth is, the more
  related are the nodes

• Combing with Google PageRank
• Some Google Results are not annotated

5
Experimental Results
6
Extending ODP Annotations To The Web

• Manual annotation for the whole web is impossible
• Biasing is an implicit way for extending
  annotations to the Web
• Is basing possible in the ODP context?
• Are ODP entries good biasing sets to obtain
  relevant results generate rankings which are
  different enough from the non-biased ranking
• When does biasing make a difference?
• Find the characteristics the biasing set has to
  exhibit in order to obtain relevant results

7
Experimental Setup

• Compare the similarity between top 100 non-biased
  PageRank results and biased results

• Similarity Measure
• OSIM degree of overlap between the top n
  elements of two rank lists

• KSim degree of agreement on ordering between
  the two rank lists

8

• Choice of Biasing Sets
• Top 0-10 PageRank pages
• Top0-2 PageRank pages
• Randomly selected pages
• Low PageRank pages
• Varied the sum of score within the set between
  0.000005 and 10 of the total sum over all pages
  (TOT).
• Experiments are done on a crawl of 3 million
  pages, and then applied on Stanford WebBase
  crawl.

9
Biasing set consists of good pages
10
Biasing set consists of random selected pages
11

• According to the random model of biasing, every
  set with TOT below 0.015 is good for biasing.
• Results are not influence by the crawl size
• (3 million crawl vs 120 million WebBase
  crawl)
• Entries in ODP have TOT below than 0.015 thus
  biasing is possible in the ODP context

12
Conclusions

• A Personalized search algorithm to rank urls
  based on the distance between user profile and
  url in the ODP taxonomy.
• Biasing on ODP entries will take effect, thus it
  is feasible to extend the manual ODP
  classification to the Web is feasible