Web Search Personalization with

snufflevoicelessInternet και Εφαρμογές Web

22 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

101 εμφανίσεις

Web Search Personalization with
Ontological User Profile

Advisor: Dr. Jai
-
Ling Koh

Speaker: Shun
-
hong Sie

Outline


Introduction


Web & Search personalization


Experimental


Conclusion

Terminology


Query


A search query that comprises of one or more
keywords.


Context


The representation of a user’s intent for
information seeking.


Ontology


Explicit specification of concepts and relationships
that can exist between them.

How do we begin search?

野草莓是怎
麼一回事?

野草莓

、圓
山、學運

For search result


Query or localized context


Domain knowledge


Long
-
term interests

Motivation


Combine search technologies and knowledge
about query and user context into a single
framework in order to provide the most
appropriate answer for a user’s information
needs.

Personalized Search


Google (or the other search engine like?!)


Feedback


Re
-
ranking


Snippets associated


Time consuming

Ontologies for web personalization


Ontological user profile


Concept are annotated by interest scores derived
and updated implicitly on the user’s information
access behavior.

Ontologies for web personalization

Cold
-
start

Match exist
concept

Collect use
behavior

Build user’s
ontology

Maintain
and update

Representation of Reference Ontology

N:total number of document in the training set

n
i

total number of documents that contain term i

n: concept

Context Model

concept

score

Spreading Activation module


Assume user behavior can be learned


Compute the weights for the relations between
each concept and all of its subconcepts using a
measure of containment





Computer a term vector for each document d
i

and compare the term vector for d
i
with the term
vectors for concept C
j

in the user profile.

Algorithm 1: Spreading Activation
Algorithm

Algorithm 2


Algorithm for the
Normalization and
Updating of Interest
Scores in the
Ontological User
Profile

Search Personalization


Algorithm 3: Re
-
ranking Algorithm

Experimental


Top
-
n Recall


Top
-
n Precision


Data Sets


ODP


Training set


test set


profile set

Experimental evaluation


User profile convergence


Rate of increase in interest scores stabilizes over
incremental updates.


Effectiveness of search personalization

User profile convergence


Average
Top
-
n Recall and Top
-
n
Precision comparisons


Standard search with

various query sizes

CONCLUSIONS


Use personal ontology can be used to
effectively tailor search results based on users’
interests and preferences.

The Norwegian National Knowledge
Base

SNL

SNL

SNL

SNL

Skien

coun
-

cil

Cap

Lex

Cap

Lex

NBL

Henrik Ibsen

Hedda
Gabler

Skien

Et dukkehjem

A doll

s house

wrote

born in

wrote


reality


topic map

information

knowledge

other topic maps

are merged in ...

Ibsen
-

centre

Ibsen
-

centre

Ibsen
-

centre

Ibsen
-

centre

Ibsen
-

centre

Ibsen
-

centre

Et dukkehjem

Helmer

Dr. Rank

Mrs. Linde

Krogstad

Nora

http://www.ontopia.net/

©

2003 Ontopia AS

Summary of Core Topic Maps Concepts

A pool of information or data


any type or format

A knowledge layer, consisting of:

knowledge layer

information layer

A
ssociations


expressing relationships
between knowledge topics

composed by

born in

composed by

O
ccurrences


information that is relevant in
some way to a given
knowledge topic

= The TAO of Topic Maps

T
opics


a set of knowledge topics for
the domain in question

Puccini

Tosca

Lucca

Madame

Butterfly