Sunteți pe pagina 1din 23

| 


   

 

Presentation by
j 

ë e core of Meta searc engine is t e result integration algorit s.

ë is presentation focuses on evaluating t e integration algorit s


of eta-searc engine, and copares eta-searc to t e general
searc engine by experients.

An experient et od to deterine t e priority of participant of


eta-searc engine is also proposed.

ë e experient result proved t at t e eta-searc engine could get


quality searc ing result on average.
|  
eneral searc engines like oogle, Ya oo, MSN etc.



As t ere are uge no. of docuents on t e world wide web, it is


very difficult to locate t e inforation t at is relevant to t e users
interest.

   
Meta Searc engine erge algorit s.

  

ë e queries posted by users w ic are subitted to t e eta


searc engine.

It sends t e query to ultiple single searc engines intern.

W en retrieved ites are returned by t e underlying searc engines


t e eta-searc engine furt er processes t ese ites and presents
relevant ites to t e user, based on a integration algorit .
|

It is very difficult to locate inforation t at is relevant to t e internet


users interest, as t ere are uge no. of docuents on t e world
wide web.

ë ere are several approac es to elp t e users to find inforation


fro t e web.

One is eta-searc et od.

Queries are subitted to t e eta-searc engine w ic sends t e


query to ultiple single searc engines.
W en retrieved ites are returned by t e underlined searc
engines furt er processes t ese ites and presents relevant ites
to t e user based on a result integration algorit .

JP calla proposed 4 typical synt esis algorit s to process


different situations.

Kirsc provided anot er typical et od w ic required t e


participant searc engine to return soe additional inforation.

ë e eta-searc engine would use t is inforation to re calculate


t e relevance of docuent on t e client side.

Meta crawler iported t e concept credibility to deterine t e


relevance between docuents and queries.
Profusions algorit  was one of t e noralized score et od.

Inquires adopted a erge strategy w ic also re calculated t e


relevance on t e client side.

I a evaluating and t en copare in experients 4 erge


algorit s t roug t is project

1. Siple erge algorit 

. abstract erge algorit 

. Position erge algorit 

4. abstract / position erge algorit 


   
ë is approac is siilar to t e single searc engine, and it just to
range t e results fro participate searc engines intern by using a
ulti wave erge et od.

In practical user are interest in searc results of t e first t ree pages


and ig probability.

ë e result after t ree pages could only en ance t e copleteness


of t e searc results but could not iprove user experience
anyore.
  

ë e ain idea of abstract erge algorit  is to rank searc


results wit t e relevance between query and t e abstract
inforation of searc results.

First, we need extract t e ters fro query, and calculate t e


relevance between ters and abstract.

Second, we calculate t e relevant between query and eac page.

Finally, t e results are returned to users according to t eir


relevance between query and abstract of one page.

  

    

ë e basic idea of position erge algorit  is to ake a full use of t e


original position inforation fro eac single searc engine.

For t e sae query, t ere are soe pages w ic will occur in several
result list of different participant searc engines, but t eir position in
different result lists ay not be t e sae.

ëo reconcile t is contradiction , we s ould take t e position in different


participants into account.
ß    

In t e position algorit  t e priority of searc engines need to be


deterined in advance.

In t is experient, I used two ain searc engines oogle and Ya oo as


t e participants.
!  

ë e abstract and t e position are very iportant inforation.


abstract/ position erge algorit  considered t ese two factors
synt etically to ake t e

integrated results to eet t e user needs.

Following is t e forulae to calculate t e relavance bertween t e


result and query.

ë e evaluation of t e results fro different erge algorit s:
In t is section, i a going to copare t e results of Siple erge
algorit , abstract erge algorit , Position erge algorit  and
abstract / position

erge algorit , t en estiates w ic result is better and ore


relevant to t e user query intent.

we use to ain t e easures accuracy, w ic is t e ratio of t e


results in t e w ole results. and t ey wait, naely t e location of
t e related results in t e w ole results.

Firstly, setting up 10 keywords w ic are ore broadly, ore


representative and ore accordant wit t e users' queries,
suc as
 eijing Olypic aes´,  eijing University 007 ill gates´,
ell Coputer´, eilongjiang University´, Jackie C an
Movie´, t e N A Yaoing´, Prison break introduction´,  ë e
aVinci Code´, ë e reat Wall´, ë inking in JAVA´.

Secondly, for eac query , we take t e first top 0 results for


t ree algorit s as t e basis data set for analyzing. Every
result will be judged w et er t e

results are related to t e keywords or not based on t e content


of eac result for eac query by uan.

If one result is relevant, we will ark t e location of t e result


w ic occur in t e 0 result. and t en calculate t e weig t of
t e eac relevant result wit t e weig t function.
j      " #
We establis ed a eta searc engine w ic take oogle and ya oo
as t e searc engines.

We used t e above 4 erge algorit s to fusion t e results fro t e


participants.

We did a test, to evaluate w et er t e eta searc engine could


iprove t e experience of t e user.

We c oose following 5 queries as t e testing set:

" ill ates eijing University 007", "eilongjiang University" , " ë e


great Wall", " eijing Olypic aes" and "ell Coputer".
Firstly, we need fetc t e top 0 results fro oogle and
Ya oo respectively for every query.

Secondly, we use t e sae way to estiate t e relevance of


t ese results respectively and ark t e positions.

ë en calculate wait for every result and accuulate t e waits.


and get t e ean value. ë en we get accuracy, (t e no of
related results) / 0
ë        
ë   " #

oogle Ya oo
Key word Accuracy Weig t Accuracy Weig t
ill ates eijing 0.6667 5.407 0.5 5.866
University 007
eilongjiang 0.667 5.4580 0. 4.994
University
ë e great Wall 0.6 5.19 0.5667 5.5166

eijing Olypic 0.4 4.6565 0.4667 5.007


aes
ell Coputer 0.5667 5.40 0.5667 5.4057

Average 0.5668 5.195 0.494 5.607


ë        
 
   " #

 

oogle 0.5467 5.195

Ya oo 0.49 5.607

Suary 0.5649 5.965

Position 0.5567 5.408

Suary/position 0.5749 5.966


j
In t is project we design an experient to evaluate abstract
erge algorit , position erge algorit , abstract/position
erge algorit  of eta searc

engine and copare t e eta- searc engine wit t e general


searc engines suc as oogle and Ya oo.

Fro t e experient results it is concluded t at t e better erge


algorit  could iprove t e quality of searc ing.

S-ar putea să vă placă și