Informetrics, Webometrics and Web Use metrics
Huimin Lu
10/21/2004
Outline
History
Article 1: Bibliometrics & WWW
Article 2: Bibliometrics of the WWW
Article 3: Authoritative Sources
Article 4: ParaSite
Conclusion
History
Term introduced by Pritchard in 1969.
Pritchard’s explanation: “the application of
mathematical and statistical methods to books
and other media of communication”.
A1: Bibliometrics and the World Wide Web
By Don Turnbull
Bibliometrics
Bibliometric laws
Apply bibliometric to WWW
Metrics design
A1: Bibliometrics
Classic citation analysis
Refined classic bibliometrics
- Standard formula for impact: n journal citations / n citable articles published
- Basic formula for immediacy index of influence:
n citations received by article during the year / total number of citable articles published
Bibliometric Coupling
- Measure the number of references two papers have in common to test for
similarity
Cocitation Analysis
- Measure the relations between cited documents
Common Errors
- multiple authors lost, self-citation, similar author names, human
error, etc.
A1: Bibliometric Laws
Bradford’s Law of Scattering
- clustering method: Ran (n from 0; a<1), sum = R/(1-a)
Lotka’s Law
- inverse square
Zipf’s Law
- familiar words with high frequency (nth word: k/n times)
A1: Applying Bibliometric to Web
Web surveys
- Georgia Tech Graphics, Visualization, and
Usability Web Surveys
Web servers
- Add programming logic
- Inaccurate data gathered: skip standard
procedures, miss state information between
usage hits, server hits themselves don’t represent
true usage.
A1: Metrics Design
Configure Web server to gather comprehensive
metrics
Manage log files
- Enhence reliability: regular backup, store log file analysis results and
logs, begin new logs timely, post results and log information for
comparasion.
- Log analysis tools: Analog, WWWStat, GetStats, Perl Scripts.
- Standardization: Extended Log File Format by WWW Consortium
Standards Committee
Downie’s attempt analysis: user-based, request,
byte-based
Optimal Web content setup & External
bibliometric gathering
A2: Bibliometrics of the World Wide Web:
An Exploratory Analysis of the Intellectual
Structure of Cyberspace
By Ray R. Larson
Analysis of 30G Web pages collected by
Inktomi “Web Crawler”
Cocitation analysis using DEC AltaVista
search engine
A2: Growth and Usage of Web
WWW
A2: Cocitation Analysis of Web
Attempt: Map the intellectual structure of Web
Question: Can cocitation techniques be
applied to charting the contents of
cyberspace?
A2: Methods
Selection of core set of items for study
Retrieval of cocitation frequency information
Compilation of the raw cocitation frequency matrix
Correlation analysis to convert the raw frequencies into
correlation coefficients
Multivariate analysis of the correlation matrix
Interpretation of the resulting “map” and validation
A2: Results
A3: Authoritative Sources in a Hyperlinked
Environment
By Jon M. Kleinberg
A new method for automatically extracting
certain types of information about a
hypermedia environment from its link
structure.
A3: Goal
Types of query search and problem
- Specific queries: scarcity problem
- Broad-topic queries: abundance problem
- Similar-page queries
Synthesize the unreliable information contained
in the presence of individual links to provide a
set of authoritative pages relevant to an initial
query.
A3: Common Approaches
Only S
- Define S to be the top k pages indexed by AltaVista
- Rank pages according to their in-degree
S -> T
- Define same root set S
- Grow S to a larger base set T
- Rank pages by their in-degree
A3: Their Approach
Extract small core sets of
community of hubs and
authorities from T
Authoritative pages
- A novel type of quality measure of the
document in hypermedia by
algorithmic means.
- Large in-degree & considerable
overlap in sets of pages that point to
them
Hub Pages
- have links to multiple relevant
authoritative pages
A3: Algorithm and Output
Method: Iteratively propagates “authority weight”
and “hub weight” across links of the web graph,
converging simultaneously to steady states for
both types of weights
Output: a pair of sets (X, Y) (X: a small set of
authorities, Y: a small set of hubs) referred by
authors as community of hubs and authorities
Claim: authoritative pages can be identified as
belonging to dense bipartite communities in the
link graph of the WWW via their algorithm.
A4: ParaSite: Mining Structural Information
on the Web
By Ellen Spertus
Varieties of link information on the Web
How the web differs from conventional
hypertext
How the links can be exploited to build useful
applications
A4: Classical Hypertext vs. Web
Classical hypertext
Web
- links don’t cross site even
document boundaries
- links can cross site and
document boundaries
- documents limited to a
single topic
- multiple topics permitted
in one web page
- manual answers each
question in exactly one
place or in none
- an answer could appear
any number of times on
the web
- Hardly change
- constantly changing
A4: Mining Links
Naïve Link Geometry
- A useful technique for finding pages on a given set of
topics
Hypertext Links
example
- Categorized into upward, downward, crosswise, and
outward
Directory Links
- Directory structure relation in pages in the absence of
hypertext links
Structure within a Page
- Page can be considered a tree of nodes, each with
attached text and links embedded in the text
Other
- Domain names, relationships between concepts
represented by words and phrases, paths traveled
through Web sites by visitors
A4: Application
Finding Moved Pages
- Exploiting hyperlinks
- Exploiting directory links
Finding Related Pages
- Collaborative filtering
- When searching for a related page with similar
pages got, ParaSite can find the page (A) that has
maximum links to the pages user got and return
other pages referneced by A.
A Person Finder
Conclusion
World Wide Web information increase
exponentially and Internet architecture turns to
be more complicated. Applying bibliometrics to
the Web will help us control and manage web
information wisely.
Example of Hypertext Link
Back to hypertext link