Relevance feedback using querylogs

1
Relevance feedback using query-logs

• Gaurav Pandey
• Supervisors
• Prof. Gerhard Weikum
• Julia Luxenburger

2
Motivation
Query
Search Engine
Results
One size fits all
3
Motivation
User info
Query
Search Engine
Results
4
Motivation
Python
5
Motivation
CGI code Debugging programming
Python
6
Usage of Query Logs
Clickthrough data

• Past queries
• Documents clicked

7
Usage of Query Logs
8
Query Reformulation
9
Language Model
Normally(without using history),
wterm d document
qquery
Importance of term w in current query
10
Language Model
Normally(without using history),
wterm d document
qquery
Importance of term w in current query
11
Language ModelHistory
Importance of term w in history instances
Importance of the term w at one instance in the
history
Documents CGI examples, program code
12
Equal Weighting
13
Discriminative Weighting
Choose different ? for every history instance..
How?
14
Overlap
if a history query has common terms
with the current query
then ?i 1, Else if there is no common term ?i0
Example Current query python information Histor
y querypython code ?i 1 History queryworld
cup ?i 0
15
Soft overlap
if a history query has common terms
with the current query
then ?i a, Else if there is no common term ?ib
(agtb)
Example Current query python information Histor
y querypython code ?i 8 History queryworld
cup ?i 2
16
Decrease with time
Use uniformly decreasing ? values If there are n
history instances, ?1 n ?2 n-1 ?3
n-2 ?n-1 2 ?n 1
17
Decrease with time
Use geometrically decreasing ? values If there
are n history instances, ?1 n ?2 n/2 ?3
n/3 ?n-1 n/(n-1) ?n 1
18
Experiment

• Comparison of the
• 4 techniques
• Equal weighting
• Basic model (without history)
• Use similar techniques for
• Probabilistic model
• Vector space model

19
Thanks