Download search engines - Eric Sieverts

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Clusterpoint wikipedia , lookup

Transcript
Eric Sieverts
University Library Utrecht
IT Department
Institute for Media &
Information Management
(Hogeschool van Amsterdam)
Google and/or/not databases
• why using search engines ?
• functionality of search engines
(including the latest technology)
• what is hidden for search engines ?
• search engines  databases
• why would people prefer google ?
• what is up for us, librarians ?
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why using search engines ?
• easy to use best match technique
• such a good relevance ranking
(at least some of them)
• still a lot of additional (hidden) functionality
• recent language technological methods
• such large collections
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why using search engines ?
some common document ranking parameters
• the more terms from your query in a document, the better
(now for most engines only "all the terms")
• the more prominent a term in a document, the better
(in <title>, in the first few sentences, in a <meta> tag)
• the more frequently repeated a search term, the better
• the closer together the terms in a document, the better
• the more uncommon a search term, the higher its weight
• the more "popular" a web-page, the better
(more hyperlinks pointing to it, more people visiting it, ..)
 google’s strong point 
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why using search engines ?
google offers a lot of additional functionality
• boolean search (if you really want to - I do occasionally!)
• "citation" search (other web-pages linking to "this" site)
• similarity search (means here: similar linking patterns;
not really better than word-based similarity search)
• disappeared documents in result set can be retrieved from
archive cache
• many other document types than just plain html
• also image search, usenet archives, integration of open
directory subject tree
see google
Eric Sieverts
|
[email protected]
see google advanced search
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why using search engines ?
modern language technology aboard
categorisation of result sets
• (formerly) northernlight's custom search folders
(rulebased method)
• teoma (statistics based method)
• wisenut (statistics based method)
• fast-alltheweb (statistics based method)
teoma
Eric Sieverts
|
[email protected]
|
wisenut
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why using search engines ?
search engine “sizes”
see for instance “search engine watch”
december 2001
search engine watch
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
what is hidden for (most) search engines ?
(and consequently for their users ! )
 non-HTML documents: flash, office-files, pdf
(not fundamentally impossible, as google demonstrates)
 "real-time" data (too difficult to keep track)
 dynamically, database generated pages
(out of fear for spider traps; but google seems to do it)
 all information hidden in searchable databases
(spiders cannot fill out database search forms)
 to-be-paid-for or licensed information
(bibliographic databases, full-text scientific journals, ....)
 all information that is not (yet) on the web
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
search engines vs. databases
besides - for us obvious - differences in content:
differences in functionality
database
search engine
 field searching
 modern retrieval technology
 boolean, proximity, truncation  relevance ranking
 controlled vocabulary
 ease of use
- categories
- thesauri
- etc
but do users use all of this ??
despite its importance !!
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
why do students graduate on google" ?
why do so many users prefer the use of search engines ?
 apparent simplicity of search engine interface
 too many separate other search systems to address
 overwhelming choice of databases
 overwhelming choice of digital primary sources
example
example
 plethora of different database system interfaces
 interfaces crowded with "functionality"
what would you use ?
– if you did't know what's the difference
– if you did't know what you'd miss
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
do you miss so much with only google ?
• google also indexes .PDF , .DOC , .PPT , .XLS , .RTF
• the web also contains preprints, reports, projects etc. that
are NOT in databases
• many scientists (and others) put copies of their published
articles on their personal websites
that seems fine, but you still get low recall, because:
• the web remains a very fragmented incomplete mess
(behind that simple google screen)
• it is not indexed consistently and in a controlled way
but for many users lousy recall is no problem at all .....
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
what is up for libraries ?
• realise better integrated access to all our precious (and
expensive) information sources
• realise more advanced retrieval possibilities while keeping
the advances of controlled indexing as well
central index solution
meta-search / portal solution
- our own choice of advanced local
search engine / retrieval software
- problems with indexing remotely
stored data
- problems with non-uniform
controlled indexing
- many remote and locally available
retrieval systems addressed in a
single query (via Z39.50, http, etc.)
- restricted to common denominator
of classical boolean functionality
- problems with non-uniform
controlled indexing
Eric Sieverts
|
[email protected]
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002
integrated system:
local central
index solution
search
central index
indexingrules for
targets
indexer
internet
full-text
links
document text files
document text files
integrated system:
metasearch /
portal solution
search
configuration
data for
targets
query-generator / result-collector
Z39.50
Z39.50
internal
api
search
search
index
index
files
files
http
internet
Z39.50
http
http xml
Z39.50
search
search
search
search
index
index
index
index
files
files
files
files
and some look into the (near) future ....
library based search systems
will improve
performance of web search
engines will improve as well
- automatic methods of uniform
classification and controlled
keyword indexing
- more flexible xml-based methods
for metasearch-solutions (srw, sru)
- improved access to remote data to
be locally indexed
- xml, rdf metadata & the semantic
web will improve concept- and
meaning- based retrieval on the web
- ever more information will be
available on the web
- newest technologies will continue to
be tested on the web first
competition between “
Eric Sieverts
|
[email protected]
“ and "our databases" will continue
|
http://www.library.uu.nl/medew/it/eric
|
Bielefeld 2002 Conference, 7 febr 2002