Transcript 2 - Department of Computer Science and Engineering, CUHK

Mining Clickthrough Data
Hao Ma
Mar 11, 2008
Definition
Hot research topics
Query Hierarchy Building
Query Suggestion
What is Clickthrough Data
Query logs recorded by search engines
2
3
Research Topics
Improving Web search ranking [E. Agichtein SIGIR 2006]
Users can help indicate most relevant results
4
Research Topics
Organize Search Results
[X. Wang, C. Zhai SIGIR 2007]
Web page summarization
[J.–T. Sun, D. Shen, H.-J. Zeng, et. al SIGIR 2005]
Query Clustering
[D. Beeferman, A. L. Berger KDD 2000]
[J.-R. We, J.-Y. Nie, H. Zhang ACM TOIS 2002]
Extraction of class attributes
[M. Pasca, B. V. Durme IJCAI 2007]
5
Mining Web Query Hierarchies from
Clickthrough Data
[D. Shen, M. Qin, W. Chen,
Q. Yang, Z. Chen AAAI 2007]
6
Mining Web Query Hierarchies from Clickthrough Data
7
Intuitions
If two queries are related to each other, they should
share some of the same or similar clicked Web pages;
For two queries qi and qj , qi is qj ’s parent if most of
the clicked pages of qj have similar pages to the
clicked pages of qi while only part of the clicked pages
of qi have similar pages to the clicked pages of qj ;
If a query is specific, the contents of its clicked pages
are relatively consistent, compared to a general query.
8
Definitions
Relative Coverage (RC)
Specificity (Spec)
9
Criterion
10
11
Demo
12
Learning Semantic Relations from
Clickthrough Data for Query Suggestion
13
Bipartite Graph
Query Similarity Graph
0.2
0.1
0.8
Queries
0.4
0.2
0.1
0.3
Websites
0.7
0.1
0.8
0.9
0.6
0.5
0.8
Given a query q,
start the
Similarity Propagation
Process
0.3
14
Similarity Propagation Model
15
Query Suggestion
16
17
18
Demo
19
Future
Event Detection
Temporal Application
Computational Linguistics
Personalization
Query Answering Search
Pay For Performance Advertising
……
20