|
Introduction:
Granted
that the current World Wide Web contains tremendous amount of information
provided by millions of users all over the world, it should be admitted
that the problem of discovering the relevant resources is not easy. The
Web has enabled users to electronically publish information accessible to
millions of people relatively easily, but as the quantity of its
information grows, the ability of those people to find relevant materials
has decreased dramatically and can be compared to looking for "a
needle in the haystack."
To solve
the problem of discovering web resources, search engines have been
developed that can provide the users with a large body of results by a
click. While the value of these tools should not be underestimated, they
have many shortcomings as information retrieval systems. With all of
their power to provide access to an enormous array of information, it has
been shown that they are finding it difficult to cope with the explosion
of web resources (Bharat & Broder, 1998; Lawrence & Giles, 1998; Bar-Ilan,
1998/99; Lawrence & Giles, 1999) and accordingly
cannot be considered as perfect tools because of their low coverage.
Their performance volatility (Rousseau, 1998/99; Snyder & Rosenbaum, 1999), fluctuations and changes in
the results set over time (Peterson, 1997; Bar-Ilan,
1998/99; Rousseau, 1999; Mettrop &
Nieuwenhuysen, 2001) and a generally low retrieval effectiveness (Gordon & Pathak, 1999) are some major shortcomings and
deficiencies which are well documented in the literature. It seems that
deficiencies of search engines in retrieving the relevant resources
mostly originate from their strategy of indexing the Web. The way in
which they index the Web, that indiscriminately harvest whatever they can
find and then do selective indexing on those contents, coupled with the
enormous mass of web resources results in overly large retrieval sets
with low relevancy. It has made it clear that without enforcement of a
more rigorous indexing strategy through some level of meta control,
search engines effectiveness and efficiency in resource discovery will
deteriorate. Since the content of the information resources has not the
right and efficient information for them to be indexed effectively, some
kind of descriptive information to impose pre-defined meaning on the Web
content is essential.
The metadata movement for resource discovery on the Web
The high
dynamics of web resources (Lawrence & Giles, 1999),
both in size and content, as well as their unique characteristics (Heery, 1996), has posed many challenges for using the
traditional procedures of resource organization and discovery, such as
cataloging rules, in the Web environment. The challenges in the way of
deploying cataloging rules for digital resources (Beacom,
2000; Huthwaite, 2001; Lagoze, 2000; Weiss and Carstens, 2001), have led to favoring
"metadata" as the best means of describing and discovering
resources on the Web.
Metadata
is a heavily loaded term for which many definitions have been offered.
It, in general, may be defined as structured data about data (Burnett,
Ng & Park, 1999, p.1212). More specifically, it is a structured
set of elements that describes the information resource for the purpose
of identification, discovery and use of information (Lee-Smeltzer,
2000, p.206). To encompass the main perspectives on metadata and
accurately reflect the current status of its studies, Burnett
et al. (1999) define metadata as "data that characterizes source
data, describes their relationships, and supports the discovery and
effective use of source data" (p. 1212).
Metadata
is a recent coinage though not a recent concept. The above definitions
about metadata are usually followed by the observation that libraries
have been producing, standardizing and maintaining metadata for a long
time; because descriptive data such as standard bibliographic
information, and indexing and cataloging information are all structured
data that describe the attributes and contents of an information resource
to facilitate their discovery and use, hence metadata. However, while the
concept of metadata is a familiar one for information professionals, in
today's jargon, this data is considered to "[be] structured so that
it can become machine-understandable as well as machine-readable […] and
has largely been identified with issues of Internet resource
discovery" (Day, 1999). As Milstead
and Feldman (1999) point out, this term "… is generally applied
to electronic resources (though it doesn't have to be) and refers to
"data" in the broadest sense--datasets, textual information,
graphics, music, and anything else that is likely to appear
electronically. While the concept includes indexing and cataloging
information (information for "resource discovery" in Webspeak),
it can go far beyond conventional document representations, such as MARC
records."
Today
metadata activities are unprecedented. Because of the exponential growth
of information resources on the Web, they expand beyond the traditional
library environment to deal with the problem of effective resource
description and discovery. The accelerated growth in the related
literature on the topic of metadata and the rapid decrease of the word
cataloging (Ercegovac, 1999) as well as several
metadata standards with different levels of richness and complexity
originated from different communities (Heery, 1996; Dempsey & Heery, 1997; Burnett, Ng &
Park, 1999), reveals the unprecedented movement towards metadata for
resource discovery on the Web.
Dublin Core Metadata Initiative: a simple metadata for the Web
Within the
diverse resource discovery activities of the mid 90's, ranging from
unstructured indexing of full-text resources by search engines to
richly-structured data like Machine Readable Cataloging (MARC) and Text
Encoding Initiative (TEI) records, Dublin Core metadata standard arose as
a means to mediate these extremes. it originated from a workshop
sponsored by Online Computer Library Center (OCLC) and the National
Center for Supercomputing Applications (NCSA) in March 1995 to "form
an international consensus on the semantics of a simple description
record for networked resources" (Weibel, Iannella
& Cathro, 1997). It was believed that resource discovery is the
most pressing need that metadata can satisfy (Weibel
et al., 1995). Therefore, only descriptive data elements required to
support resource discovery were considered and data elements covering
other characteristics of the resource such as terms and conditions,
archival status, and other types of metadata were not included (Dempsey & Weibel, 1996).
The
primary deliverable from the OCLC/NCSA workshop was a set of thirteen
metadata elements, named the Dublin Core Metadata Element Set (or Dublin
Core, for short) by the workshop participants. The Dublin Core was
proposed as the minimum number of metadata elements required to
facilitate the resource discovery in a networked environment such as the
Internet (Weibel et al., 1995); and until the third
workshop, the elements were increased to 15 (Weibel &
Miller, 1997). This metadata elements set includes Title, Creator,
Subject, Description, Publisher, Contributor, Date, Type, Format,
Identifier, Source, Language, Relation, Coverage, and Rights (Dublin Core, 1999).
The
functions of the Dublin Core elements can be categorized into four
classes, according to the four elementary uses of the bibliographic data.
The IFLA statement on the purpose of bibliographic records identifies
four 'generic tasks' the users perform and these records should support (IFLA Study Group on the Functional Requirements for
Bibliographic Records, 1998): To find, to identify, to select, and to
acquire or obtain the resource. Dublin Core metadata elements support
these four generic tasks as follows (Dublin Core Metadata
and the Cataloging Rules, 1998):
· FIND:
Elements that are likely to be primary search categories for discovery or
finding of electronic resources include Title, Creator, Contributor, and
Subject. The elements that are likely to be secondary or restricting
features of a search are Language, Coverage, and Format.
· IDENTIFY: the elements that are related to the "instantiation"
of the resource include Date, Type, Format, and Identifier.
· SELECT: the elements intended to provide some information for a user to
make a selection among multiple search results include Description and
Coverage.
· OBTAIN: In a networked environment, obtaining a resource should be
fully supported by the inclusion of an accurate address in the Identifier
element.
The
implementation of Dublin Core elements on the Web requires a formal
syntax. In 1996, a consensus concerning embedding metadata in HTML was
reached at the W3C Distributed
Indexing and Searching Workshop (Dempsey &
Weibel, 1996). Because of the changes in HTML as well as a general
need for greater formalization of the syntax, an Internet Draft authored
by John Kunze (1999) was released after the debates at
the sixth Dublin Core workshop, which explains how to encode Dublin Core
elements in HTML. Current implementation of Dublin Core on the Web is
often based on metadata embedded in HTML metatags.
Metadata Effectiveness: problem statement
With any
metadata schema, there is a question of effectiveness. Does metadata
provide a basis for increased effectiveness of retrieval by search
engines? While there have been many studies done to evaluate search
engines from different points of view, few studies have been done to test
the effectiveness of metadata on resource discovery by search engines. Turner and Brackbill (1998) did a research on how the
embedded metadata (HTML metatags) effects retrieval of web pages. The use
of keywords metatag was shown to cause the significant improvement in the
retrievability of a web page. However, another type of metatag
(description metatag) exhibited no improvement in retrieval. Henshaw
and Valauskas (2001) studied the effectiveness of Dublin Core
metadata together with HTML keywords and description metatags on
enhancing information retrieval in a suit of specific search engines.
Results suggested that metadata did not play a significant role in
increasing the likelihood of a web page being indexed or highly ranked by
search engines.
This study
aims to examine the following questions related to the use of four Dublin
Core metadata elements, which are likely to be primary search categories
for resource discovery:
· Do Dublin Core elements (Title, Subject,
Creator and Contributor) improve the retrieval rank of a web page?
· Is retrieval performance of the major search
engines improved after embedding metadata into the web pages?
Methodology
The web pages
tested in this study, are a group of articles published in the home page
of the Iranian International Journal
of Science (freely available at:
http://www.fos.ut.ac.ir/~journal/iijs.html). At the time of this research,
the total number of the articles published online by this journal was 16
articles (see Appendix A). The articles were
submitted to the major search engines (see table 1). Among these search
engines, AOL Search, HotBot and Iwon failed to index the submitted
articles and AlltheWeb, AltaVista, Google, Lycos, MSN Search, Excite and
WebCrawler indexed the articles. Table 1 shows the results of searches of
the titles in the search engines. The presence of the articles in the
databases of the search engines is indicated by "+".
As table 1 indicates, MSN indexed
only 4 articles and excluded from the study. Therefore, the maximum
number of articles indexed by the maximum number of search engines is 10
and 6 respectively. These articles are shown by "*" sign. Table
2 shows the search engines that have indexed the articles and are tested
in this study.
Table 1. The presence of articles in the database of Internet
search engines
|
Article
|
Allthe Web
|
Alta Vista
|
AOL Search
|
Google
|
HotBot
|
Iwon
|
Lycos
|
MSN Search
|
Excite
|
Web Crawler
|
|
1
|
-
|
+
|
-
|
+
|
-
|
-
|
-
|
+
|
-
|
+
|
|
2
|
-
|
+
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
+
|
|
3
|
-
|
+
|
-
|
-
|
-
|
-
|
-
|
-
|
-
|
+
|
|
4
|
-
|
+
|
-
|
+
|
-
|
-
|
-
|
+
|
-
|
+
|
|
5*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
6
|
-
|
+
|
-
|
+
|
-
|
-
|
-
|
-
|
+
|
+
|
|
7*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
8*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
+
|
+
|
+
|
|
9*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
10*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
+
|
+
|
+
|
|
11*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
12*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
13*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
14*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
15*
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
+
|
+
|
|
16
|
+
|
+
|
-
|
+
|
-
|
-
|
+
|
-
|
-
|
-
|
Table 2. Search engines tested in this study
|
Search Engine
|
URL
|
|
AlltheWeb
|
www.alltheweb.com
|
|
AltaVista
|
www.altavista.com
|
|
Excite
|
www.excite.com
|
|
Google
|
www.google.com
|
|
Lycos
|
www.lycos.com
|
|
WebCrawler
|
www.webcrawler.com
|
In the
next step, keywords were extracted from the articles. Keyword extraction
was performed in accordance with their corresponding Dublin Core elements.
All of the keywords in the titles were extracted as the value of element
"title"; the keywords provided by the authors in each article
were considered as the value of element "subject"; the first
creator of each article was considered as the value of element
"creator" and the other ones (if any) as the value of element
"contributor". In the case of some articles in which there was
a complete overlap between subject and title keywords (all of the
keywords assigned as subject keyword by authors existed in the title), a
keyword was extracted from the abstract as the value of element subject
to avoid any common keywords between title and subject elements. Totally
82 keywords were extracted from articles.
Using the
simple search of search engines and regarding the nature of searching in
each of engines, the keywords were searched. The phrases were searched
with double quotes, so that the entire phrase was searched rather than
each word of the phrase. The retrieval rank of a web page in a search
engine results list was used to measure performance of the web pages. The
higher the retrieval rank of a web page in a search engine results list,
the better its performance and vice versa.
As at the
first Text Retrieval Conference,
using 200 results was reported as a retrieval threshold (Turner
& Brackbill, 1998, p.264), the first 200 results of each search
were examined as an arbitrary cutoff point and the ranking of the first
specific reference to the exact web page within those first 200 hits was
recorded. If a search could have resulted in retrieving a page but it was
not in the top 200 results, that keyword was given the rank of 201 for
that search. Rankings, therefore, ranged from 1 (highest) to 201 (not
retrieved).
In the
next step, the web pages were randomly divided into two groups:
experimental group and control group. It resulted in totally 43 keywords
in experimental group and 39 keywords in the control group. The metadata
elements were embedded into the web pages of experimental group through
HTML metatags. Figure 1 shows the metadata elements of one of the web
pages.
To ensure
that search engines revisit the pages, a numeric character (digit 1) was
added to the beginning of the titles of the pages in HTML title tag.
Since search engines take the title of each page from this tag and show
it to the user in the list of the retrieved results, showing the changed
title (a title with digit 1) was considered as an indicator for ensuring
that the search engines had revisited the pages through their continuous
crawling and refreshing. It took 4 months for search engines to revisit
all pages and among them, Google was the first one and AltaVista was the
last one that revisited all pages. From the perspective of a content
provider and regarding the high dynamics of the Web, it is highly
suggested that the speed of revisiting web pages and updating the
databases by Internet search engines might be improved.
Figure 1. Dublin Core elements in HTML metatags
|
<title> 1 Theoretical and Experimental
Investigation on Back-scattered Low Energy Gamma Radiation from
Different Metals <title/>
<META NAME="DC.Title" CONTENT ="Theoretical and
Experimental Investigation on Back-scattered Low Energy Gamma Radiation
from Different Metals">
<META NAME="DC.Creator" CONTENT="A.
Pazirandeh">
<META NAME="DC.Subject" CONTENT="Compton scattering,
Rayleigh scattering, Double scattering, Albedo spectrum, Coherent and
incoherent scattering, photo-electric effect">
<META NAME= "DC.CONTRIBUTOR" CONTENT="N. Sobhkhiz">
|
The web
pages were controlled in order not to be changed during the study. The
only change was metadata implementation in the experimental group. Once
search engines revisited the pages, the searches were repeated with the
same search terms and exactly in the same fashion and the results were
recorded for future comparison. Appendix B and C indicate the first (R1) and second (R2) ranking of
the keywords in the experimental and control groups. To determine the
differences between the first and the second ranks, the first rank of
each keyword was subtracted from its second rank. The differences
achieved are indicated in R3 column. The negative numbers indicate losing
ground and positive ones indicate gaining ground in the ranking of
keywords.
Analysis of
Findings
To
determine whether metadata elements have affected the retrieval rank of
the web pages, the achieved ranks (R3) of the keywords in the
experimental and control groups were compared. The comparison was made by
running Mann-Whitney U test. The U statistic "is used to test the
significance of differences in central tendency between independent
groups when the scores are ranks or when ranks have been substituted for
the original scores" (Willemsen, 1974, p.193).
This provides for a comparison of two sets of ranked scores and tests
rankings of at least rank-order or ordinal level data. Since the results
provided by the search engines are considered ordinal level data (Turner & Brackbill, 1998, p.265), Mann Whitney U test
can be used to determine the significance of differences of ranks between
two independent groups in this study. This test was run by SPSS
(Statistical Package for Social Sciences) software.
In order
to answer the question that whether Dublin Core elements improve the
retrieval rank of a web page, the R3 of the keywords in experimental and
control groups were compared. R3 of two groups were compared to determine
if there is a significant difference for two groups with respect to the
web pages retrieval ranks before and after the metadata implementation.
Table 3 and 4 represent the descriptive statistics and Mann-Whitney U test
statistics of the comparison.
Table 3. Descriptive Statistics of experimental and control groups
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
258
|
-187
|
200
|
9.736
|
52.74
|
|
Control
|
234
|
-200
|
200
|
9.974
|
51.05
|
Table 4. Mann-Whitney U Test Statistics of experimental and
control groups
|
Mann-Whitney U
|
29437.000
|
|
Z
|
-.530
|
|
Asymp. Sig.
(2-tailed)
|
.596
|
The
significance level (p) or sig. for all tests is 0.01. Adopting .01 or 1 per
cent as the significance level at or below which the difference of ranks
between control and experimental groups are unlikely due to the chance,
we can determine whether these differences are statistically significant
or not. That is, if P .01 then we can conclude that the differences
between two groups are statistically significant. As table 4 indicates,
the significance level (P=596) is greater than .01 (.596>.01). The R3
Mean of the web pages with metadata elements and those without metadata
elements are 9.73 and 9.94 respectively. Therefore, there is no
statistically significant difference between experimental and control
groups with respect to their retrieval rank improvement. In other words,
using Dublin Core elements (Title, Subject, Creator and contributor) did
not affected the retrieval rank of the web pages.
Is
retrieval performance of the major search engines improved after
embedding metadata into the web pages? To answer the second question of
this study, each search engine was considered separately. As search
engines use different algorithms for indexing and ranking the web pages,
to determine the significance of the differences of ranks between
experimental and control groups, the U test was run for each search
engine.
Tables 5
and 6 show the statistical results for AlltheWeb. From table 6, the
differences between two groups are not significant, (P=.410 [>0.01]).
In other words, there is no statistically significant difference between
the experimental and control groups and therefore, the retrieval
performance of AlltheWeb is not improved after metadata implementation.
The R3 Mean for the web pages with metadata and those without metadata
are -2.55 and -2.84 respectively (see table 5). This, therefore, shows
that the web pages with metadata did not achieve better rankings than the
web pages without metadata in AlltheWeb.
Table 5. Descriptive Statistics for AlltheWeb
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-94
|
86
|
-2.55
|
27.44
|
|
Control
|
39
|
-94
|
74
|
-2.84
|
21.21
|
Table 6. Mann-Whitney U Test Statistics for AlltheWeb
|
Mann-Whitney U
|
785.500
|
|
Z
|
-.825
|
|
Asymp. Sig.
(2-tailed)
|
.410
|
The
statistical results for AltaVista are presented in the tables 7 and 8. It
is evident that the differences between the pages in the experimental and
control groups with respect to their retrieval rankings are not
statistically significant because of P=.519 (>0.01). It suggests that
the variances of the rankings for two groups are not substantially
different. The R3 Mean ranks for the experimental group and the control
group are -2.51 and -7.10 respectively. Therefore, we cannot assume that
the metadata implementation has enhanced web pages retrievability and
ranking in AltaVista.
Table 7. Descriptive Statistics for AltaVista
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-39
|
2
|
-2.51
|
7.31
|
|
Control
|
39
|
-198
|
2
|
-7.10
|
31.97
|
Table 8. Mann-Whitney U Test Statistics for AltaVista
|
Mann-Whitney U
|
776.50
|
|
Z
|
-.644
|
|
Asymp. Sig.
(2-tailed)
|
.519
|
Tables 9
and 10 demonstrate the statistical results for Excite. From tables, the
web pages in the experimental group (R3 Mean =27) did not achieve the better
ranking than the web pages in the control group (R3 Mean=32.38) because
of P=.729 (>0.01). It suggests that the variances of the rankings for
two groups are not substantially different. The web pages with metadata
elements, therefore, did not achieve better performance than those
without metadata elements with respect to their retrieval ranking in
Excite.
Table 9. Descriptive Statistics for Excite
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-187
|
200
|
27
|
85.04
|
|
Control
|
39
|
-9
|
200
|
32.38
|
69.26
|
Table 10. Mann-Whitney U Test Statistics for Excite
|
Mann-Whitney U
|
806
|
|
Z
|
-.346
|
|
Asymp. Sig.
(2-tailed)
|
.729
|
Tables 11 and
12 show the statistical results for Google. From table 12, the
differences between the rankings of the web pages in the experimental
group and the control group are not statistically significant (P=188
[>0.01]). The R3 Mean for the web pages with metadata elements and
those without metadata elements are 12.25 and 7.64 respectively.
Therefore, we cannot assume that metadata elements have caused better
performance for the web pages with respect to their retrieval ranks in
Google.
Table 11. Descriptive Statistics for Google
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-90
|
189
|
12.25
|
32.71
|
|
Control
|
39
|
-21
|
159
|
7.64
|
29.33
|
Table 12. Mann-Whitney U Test Statistics for Google
|
Mann-Whitney U
|
707.50
|
|
Z
|
-1.31
|
|
Asymp. Sig.
(2-tailed)
|
.188
|
The
statistical results for Lycos are demonstrated in the tables 13 and 14.
It is evident that the differences between the web pages in the experimental
group (R3 Mean=1.69) and the control group (R3 Mean=-2.25) with respect
to their retrieval rankings are not statistically significant because of
P=.434 (>0.01). It suggests that the variances of the rankings for two
groups are not substantially different. Therefore, we cannot assume that
the metadata implementation has enhanced web pages retrievability and
ranking in Lycos.
Table 13. Descriptive Statistics for Lycos
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-70
|
84
|
1.69
|
22.05
|
|
Control
|
39
|
-92
|
75
|
-2.25
|
20.81
|
Table 14. Mann-Whitney U Test Statistics for Lycos
|
Mann-Whitney U
|
762
|
|
Z
|
-.783
|
|
Asymp. Sig.
(2-tailed)
|
.434
|
Tables 15 and
16 show the statistical results for WebCrawler. From tables, the
differences of rankings between the web pages in the experimental group
(R3 Mean=22.53) and the control group (R3 Mean= 32.02) are not
statistically significant (P=.547 [>0.01]). Metadata elements,
therefore, have not caused better performance for the web pages with
respect to their ranking and retreivability in WebCrawler.
Table 15. Descriptive Statistics for WebCrawler
|
Groups
|
N
|
Minimum
|
Maximum
|
R3 Mean
|
Std. Deviation
|
|
Experimental
|
43
|
-187
|
200
|
22.53
|
81.08
|
|
Control
|
39
|
-200
|
200
|
32.02
|
82.31
|
Table 16. Mann-Whitney U Test Statistics for WebCrawler
|
Mann-Whitney U
|
781.50
|
|
Z
|
-.602
|
|
Asymp. Sig.
(2-tailed)
|
.547
|
Table 16 shows
the significance level (P) of search engines. The significance level of
every search engine is more than .01 and consequently, it suggests no
statistically significant difference between the ranks of the pages with
Dublin Core elements and the pages without Dublin Core elements. In other
words, it is unreasonable to assume that the use of four Dublin Core
metadata elements has led to an improvement in the retrieving and ranking
of the web pages through six search engines: AlltheWeb, AltaVista, Excite,
Google, Lycos and WebCrawler.
Table 17. Mann-Whitney U Test Results for Search Engines
|
Search engines
|
Significance
Level (P=)
|
|
AlltheWeb
|
.410
|
|
AltaVista
|
.519
|
|
Excite
|
.729
|
|
Google
|
.188
|
|
Lycos
|
.434
|
|
WebCrawler
|
.547
|
Conclusion
The
current strategy of search engines to indiscriminately harvest whatever
they can find and then do full text indexing on those contents is an
unsustainable for resource discovery and generally results in low relevancy
in retrieval. Therefore, providing descriptive data to impose some level
of meta control on the content is necessary for the current Web to be
more effective and efficient. One of the best solutions to web resource
discovery is the embedding of descriptive metadata in Web for harvesting
by web index services. It consequently has led to developing and
maintaining descriptive metadata schemas. Among the current metadata
standards, Dublin Core has the potential of being adapted as an
international standard for resource description and discovery on the Web
and as a lingua franca
for metadata. The goal of this study was to determine whether the Dubline
Core implementation could improve web resource discovery via search
engines. The effectiveness of four Dublin Core elements that concentrate
on resource discovery was evaluated including Title, Subject, Creator,
and Contributor. Two questions were considered in this study: Do Dublin
Core elements improve the retrieval rank of a web page? and Is retrieval
performance of the major search engines improved after embedding metadata
into the web pages? Towards these aims, the articles published online by
the Iranian International Journal of
Science in the form of HTML pages (16 articles at the time of
study) were considered as testing web pages and were submitted to 10
major search engines. The maximum number of search engines that indexed
the maximum number of articles was 6 and 10 respectively. Keywords
extracted from the indexed web pages were searched in the 6 search
engines and their retrieval ranks were recorded for future comparisons.
Dividing the web pages into two experimental and control groups, the
metadata elements were embedded into the web pages of experimental group.
After that search engines revisited the pages through their continuous
crawling and refreshing, the searches were repeated exactly in the same
way and the results were recorded. Mann-Whitney U Test was employed to
compare the results and examine two questions.
Based on the
statistical analysis discussed in the previous section, and regarding the
first question of the present study, it was found that using Dublin Core
elements did not improve the retrieval rank of the web pages.
Mann-Whitney U test comparisons of rankings of pages with metadata
elements versus those without metadata elements did not reveal a
statistically significant difference at the .01 level. The lack of a
significant difference between two groups of web pages shows that four
Dublin Core elements do not affect the retreivability and ranking of web
pages and consequently is not an impact factor for resource discovery on
the current Web. To answer the second question of the study, the
retrieval performance of 6 search engines (AlltheWeb, AltaVista, Google,
Lycos, Excite and WebCrawler) before and after metadata implementation
was examined. Final statistical analysis revealed that the difference of
ranks of the pages with metadata and those without metadata in each
search engine was not significant and thus the retrieval performance of
none of the search engines improved after metadata use. It shows that
Dublin Core metadata, as a well-known metadata schema, is not widely
accepted and used by search engine designers and the spiders do not
consider its elements while ranking the web pages.
Resource
discovery is impossible without resource description and adequate
resource description assures effective discovery (Dillon,
2001). It is believed that the greatest potential for improvements to the
resource discovery on the Web lies in the use of metadata. Undoubtedly,
there is value in the current search engines as the main resource
discovery tools on the Web, which operate without the aid of descriptive
metadata. However, for them to be more effective and efficient metadata
has to matter and they have to move beyond the full text indexing of the
Web. Creating the metadata schemas for web resources is essential, but
not sufficient. For a metadata schema to be an impact factor in resource
discovery, it has to be widely accepted and deployed both by content
providers and by web indexing services in a systematic way. As Lynch (2001, p.14) asks, if web indexing services do not
use metadata, who will go to the expense and trouble of creating and
maintaining it?
Acknowledgment
The Author would like to express his
special gratitude to Prof. Abbas Horri for his encouragements and
assistances. The author also would like to give thanks to Mr. Keyvan
Salehi who helped with the statistical analysis and Mr. Darush
Alimohammadi for his ongoing and useful debates on the subject.
References
·
Bar-Ilan, J. (1998/99). Search
engine results over time: A case study on search engine Stability. Cybermetrics, 2/3, (1). Retrieved
January 8, 2004, from
http://www.cindoc.csic.es/cybermetrics/articles/v2i1p1.html
·
Beacom, M. (2000). Crossing a
digital divide: AACR2 and unaddressed problems of networked resources.
Retrieved November 27, 2004, from
http://www.loc.gov/catdir/bibcontrol/patton_paper.html
·
Bharat, K. & Broder, A.
(1998). A technique for measuring the relative size and overlap of public
web search engines. Computer
Networks and ISDN Systems, 30, 379-388.
·
Burnett, K., Ng, K. & Park,
S. (1999). A comparison of the two traditions of metadata developments. Journal of the American Society for Information
Science, 50(13), 1209-1217.
·
Day, M. (1999). Metadata
and electronic information. Retrieved December 1, 2004, from
http://www.ukoln.ac.uk/metadata/presentations/circe/birmingham.html
·
Dempsey, L. & Heery, R.
(1997). A
review of metadata: A survey of content resource description formats.
Retrieved December 1, 2004, from http://www.ukoln.ac.kr/metadata/DESIRE/overview/rev_it.html
·
Dempsey, L. & Weibel, S. L.
(1996). The
Warwick metadata workshop: A framework for the development of resource
description. D-Lib Magazine,
(July/August). Retrieved July 3, 2004, from
http://www.dlib.org/dlib/july96/07weibel.html
·
Dillon, M. (2001). Metadata
for web resources: how metadata works on the Web. Retrieved July 3,
2004, from http://www.loc.gov/catdir/bibcontrol/dillon-paper.html
·
Dublin
Core metadata and the cataloging rules (1998). Retrieved December 1,
2004, from
http://www.libraries.psu.edu/iasweb/personal/jca/dublin/dcreport.html
·
Dublin Core
metadata element set, version 1, 1: reference description (1999).
Retrieved November 26, 2004, from
http://dublincore.org/documents/1999/07/02/dces
·
Ercegovac, Z. (1999).
Introduction. Journal of American
Society for Information Science, 50(13), 1165-1168.
·
Gordon M. & Pathak P. (1999).
Finding information on the World Wide Web: the retrieval effectiveness of
search engines. Information Processing and Management, 35(2), 141-180.
·
Heery, R. (1996). Review of
metadata formats. Program,
30(4), 345-373.
·
Henshaw, R. & Valauskas, E.
J. (2001). Metadata as a catalyst: experiments with metadata and search
engines in the internet journal, First Monday. Libri, 51(2), 86-101.
·
Huthwaite, A. (2001). AACR2
and its place in the digital world: near-term solutions and long-term
direction. Retrieved November 27, 2004, from
http://www.loc.gov/catdir/bibcontrol/huthwaite_paper.html
·
IFLA study group on the
functional requirements for bibliographic records (1998). Functional requirements
for bibliographic records: Final report. Retrieved December 1, 2004,
from http://www.ifla.org/VII/s13/frbr/frbr.htm
·
Kunze, J. (1999). Encoding Dublin Core
metadata in HTML. Retrieved December 4, 2004, from
http://www.fuqs.org/rfcs/rfcs2731/html
·
Lagoze, C. (2000). Business
unusual: how "event-awareness" may breathe life into the
catalog? Retrieved November 27, 2004, from
http://www.cs.cornell.edu/lagoze/papers/lagozelc.pdf
·
Lawrence, S. & Giles, C. L.
(1998). Searching the World Wide Web. Science,
280, 98-100.
·
Lawrence, S. & Giles, C. L.
(1999). Accessibility and distribution of information on the Web. Nature, 400, 107-110.
·
Lee-Smeltzer, K. (2000). Finding
the needle: controlled vocabularies, resource discovery and dublin core. Library Collections, Acquisitions &
Technical Services, 24, 205-215.
·
Lynch, C. A. (2001). When
documents deceive: trust and provence
as new factors for information retrieval in a tangled web. Journal of the American Society for
Information Science and Technology , 52(1), 12-17.
·
Mettrop, W. & Nieuwenhuysen,
P. (2001). Internet search engines - fluctuations in document
accessibility. Journal of
documentation, 57 (5), 623-651.
·
Milstead, J. & Feldman, S.
(1999). Metadata: Cataloging
by any other name. originally published in ONLINE, January 1999, Retrieved
December 19, 2004 from: http://www.cbuc.es/5digital/1.pdf
·
Peterson, R. E. (1997). Eight
Internet search engines compared. First
Monday, 2(2). Retrieved December 20, 2004, from
http://www.firstmonday.dk/issues/issue2_2/peterson/index.html
·
Rousseau, R. (1998/99). Daily
time series of common single word searches in AltaVista and NorthernLight.
Cybermetrics, 2/3(1).
Retrieved December 15, 2004, from
http://www.cindoc.csic.es/cybermetrics/articles/v2i1p2.html
·
Rousseau, R. (1999). Time
evolution of the number of hits in keyword searches on the Internet.
Post Conference Seminar - Cybermetrics'99 at the Seventh International Conference on
Scientometrics and Informetrics, July 9, 1999, Colima, Mexico.
Summary Retrieved December 27, 2004, from
http://www.cindoc.csic.es/cybermetrics/cybermetrics99.html
·
Snyder, H. & Rosenbaum, H.
(1999). Can search engines be used as tools for web-link analysis? A
critical view. Journal of
Documentation, 55(4), 375–384.
·
Turner, T. P. & Brackbill, L.
(1998). Rising to the top: evaluating the use of HTML metatag to improve
retrieval of World Wide Web documents through Internet search engines. Library Resources and Technical Services,
42 (4), 258-271.
·
Weibel, S., Godby, J., Miller, E.
& Daniel, R. (1995). OCLC/NCSA metadata workshop report.
Retrieved November 26, 2004, from http://www.ifla.org/documents/libraries/cataloging/oclcmeta.htm
·
Weibel, S., Iannella, R. &
Cathro, W. (1997). The 4th
Dublin Core metadata report: DC-4, March 3-5, 1997, National Library
of Australia, Canberra.
D-Lib Magazine, (June).
Retrieved November 26, 2004, from
http://www.dlib.org/dlib/june97/metadata/06weibel.html
·
Weibel, S. & Miller, E.
(1997). Image
description on the Internet: A summary of the CNI/OCLC image metadata
workshop, September 24-25, 1996, Dublin, Ohio.
D-Lib Magazine,
(January). Retrieved November 26, 2004, from
http://www.dlib.org/dlib/january97/oclc/01weibel.html
·
Weiss, A. K. & Carstens, T.
V. (2001). The year’s work in cataloging, 1999. Library Resources and Technical Services , 45(1), 47-53.
·
Willemsen, E.W. (1974). Understanding statistical reasoning: how to
evaluate research literature in the behavioral sciences. San Francisco: W.H.
Freeman and Company.
Appendix A: The articles' titles of the Iranian International
Journal of Science
|
No.
|
Title
|
|
1
|
Some Production
Comparisons of Two Celluloytic Fungi
|
|
2
|
Third Virial Coefficient
and Compressibility Factors for Dense Spherical Gases Using the HFD-C
Potential
|
|
3
|
Digenetic
Studies, a Key to Reveal the Timing of Oil Migration: an Example from
the Tirrawarra Sandston Reservoir, Southern Cooper
Basin, Australia
|
|
4
|
Optimal Control
of an Inhomogeneous Problem by Using Measure Theory
|
|
5
|
Probe Diagnostics
of Confined Plasma Produced by 13.56 MHz R.F Plasma Source
|
|
6
|
Occurrence and
Distribution of Aquatic saprolegniaceae in Northwest and South of
Tehran
|
|
7
|
Effect of Pectic
Acid and b-Glocan on Prolactin Secretion by Ovine Pituitary Explants
|
|
8
|
Deformational
Behavior of Quartz and Feldspar in Quartzites within Shear Zones in the
Adelaide Hills Area, South
Australia
|
|
9
|
Construction of some
Join Spaces Boolean Algebras
|
|
10
|
Characterization
of Certain Infinitely Divisible Distributions
|
|
11
|
Theoretical and
Experimental Investigation on Back-Scattered Low Energy Gamma Radiation
from Different Metals
|
|
12
|
Cytogenetic Biomonitoring
of Workers Occupationally Exposed to Aromatic Solvents
|
|
13
|
Notes on the
Distribution, Climate and Flora of the Oil Field Areas, South-West of Iran
|
|
14
|
Isotopic
Signature of the Diagenetic Fluids and Cement in the Tortachilla Limestone, South
Australia
|
|
15
|
Correlating
marine Palynomorph Variations with Sequence Boundaries of Upper
Jurassic Sediments in a Basin
of Northern Switzerland
|
|
16
|
On Approximately
Convex Functions
|
Appendix B: The rankings of the keywords in the experimental group
|
|
|
AlltheWeb
|
AltaVista
|
Excite
|
Google
|
Lycos
|
WebCrawler
|
|
Element
|
Keyword
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
|
Title
|
Energy Gamma
Radiation
|
40
|
36
|
+4
|
29
|
29
|
0
|
201
|
201
|
0
|
178
|
136
|
+42
|
40
|
36
|
+4
|
201
|
201
|
0
|
|
Title
|
Metals
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Pectic Acid
|
1
|
4
|
-3
|
6
|
7
|
-1
|
1
|
1
|
0
|
7
|
6
|
+1
|
1
|
4
|
-3
|
1
|
1
|
0
|
|
Title
|
b-Glucan
|
5
|
9
|
-4
|
26
|
30
|
-4
|
15
|
30
|
-15
|
74
|
53
|
+21
|
5
|
9
|
-4
|
15
|
29
|
-14
|
|
Title
|
Prolactin Secretion
|
3
|
23
|
-20
|
10
|
10
|
0
|
14
|
201
|
-187
|
160
|
92
|
+68
|
3
|
23
|
-20
|
14
|
201
|
-187
|
|
Title
|
Ovine Pituitary
Explants
|
2
|
1
|
+1
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
2
|
1
|
+1
|
1
|
1
|
0
|
|
Title
|
Deformational
Behavior
|
1
|
1
|
0
|
8
|
6
|
+2
|
201
|
1
|
+200
|
1
|
1
|
0
|
1
|
1
|
0
|
201
|
1
|
+200
|
|
Title
|
Quartz
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Feldspar
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
153
|
+48
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Quartzites
|
96
|
82
|
+14
|
162
|
201
|
-39
|
201
|
55
|
+ 146
|
201
|
12
|
+189
|
97
|
82
|
+15
|
201
|
57
|
+144
|
|
Title
|
Shear Zones
|
201
|
201
|
0
|
165
|
187
|
-22
|
201
|
40
|
+161
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
30
|
+171
|
|
Title
|
Adelaide Hills
Area
|
101
|
15
|
+86
|
1
|
6
|
-5
|
201
|
39
|
+162
|
201
|
201
|
0
|
99
|
15
|
+84
|
201
|
45
|
+156
|
|
Title
|
South Australia
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Probe Diagnostics
|
102
|
175
|
-73
|
11
|
10
|
+1
|
201
|
201
|
0
|
88
|
54
|
+34
|
127
|
175
|
+48
|
201
|
201
|
0
|
|
Title
|
Confined Plasma
|
201
|
201
|
0
|
23
|
26
|
-3
|
201
|
201
|
0
|
166
|
103
|
+63
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
R.F Plasma Source
|
1
|
1
|
0
|
1
|
1
|
0
|
201
|
7
|
+194
|
59
|
44
|
+15
|
1
|
1
|
0
|
201
|
8
|
0
|
|
Title
|
Isotopic
Signature
|
100
|
194
|
-94
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
100
|
144
|
-44
|
201
|
201
|
0
|
|
Title
|
Diagenetic fluids
|
69
|
10
|
+59
|
201
|
201
|
0
|
22
|
201
|
-179
|
55
|
29
|
+26
|
70
|
10
|
+60
|
22
|
201
|
-179
|
|
Title
|
Cement
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Tortachilla
Limestone
|
7
|
1
|
+6
|
201
|
201
|
0
|
1
|
1
|
0
|
4
|
3
|
+1
|
7
|
1
|
+6
|
1
|
1
|
0
|
|
Title
|
South Australia
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Compton scattering
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Reyleig scattering
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Double scattering
|
82
|
160
|
-78
|
22
|
27
|
-5
|
201
|
57
|
+144
|
49
|
29
|
+20
|
83
|
153
|
-70
|
201
|
58
|
+143
|
|
Subject
|
Albedo spectrum
|
12
|
7
|
+5
|
1
|
1
|
0
|
201
|
1
|
+200
|
10
|
5
|
+5
|
12
|
9
|
+3
|
201
|
1
|
+200
|
|
Subject
|
Coherent and
incoherent scattering
|
54
|
62
|
-8
|
18
|
26
|
-8
|
201
|
55
|
+146
|
39
|
38
|
+1
|
54
|
58
|
-4
|
201
|
55
|
+146
|
|
Subject
|
Photo-electric
effect
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Plant extracts
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Intracrystalline
deformation
|
4
|
3
|
+1
|
3
|
3
|
0
|
6
|
10
|
-4
|
14
|
11
|
+3
|
4
|
3
|
+1
|
6
|
10
|
-4
|
|
Subject
|
R.F Plasma
reactor
|
3
|
8
|
-5
|
4
|
4
|
0
|
201
|
15
|
+186
|
27
|
25
|
+2
|
4
|
8
|
-4
|
201
|
15
|
+186
|
|
Subject
|
Stable isotopes
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Meteoric cement
|
1
|
1
|
0
|
1
|
3
|
-2
|
5
|
2
|
+3
|
5
|
6
|
+1
|
1
|
1
|
0
|
5
|
2
|
+3
|
|
Subject
|
Diagenesis
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Isotopic
composition
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Creator
|
A. Pazirandeh
|
2
|
2
|
0
|
201
|
201
|
0
|
3
|
9
|
-6
|
201
|
201
|
0
|
2
|
2
|
0
|
3
|
9
|
-6
|
|
Creator
|
Houri Sepehri
|
1
|
1
|
0
|
1
|
2
|
-1
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Creator
|
Ali Yassaghi
|
2
|
3
|
-1
|
1
|
1
|
0
|
2
|
2
|
0
|
4
|
7
|
-3
|
2
|
3
|
-1
|
2
|
2
|
0
|
|
Creator
|
M. Khorassani
|
1
|
2
|
-1
|
2
|
2
|
0
|
9
|
1
|
+8
|
1
|
1
|
0
|
1
|
2
|
-1
|
9
|
1
|
+8
|
|
Creator
|
Hossain
Rahimpour-Bonab
|
5
|
6
|
-1
|
4
|
5
|
-1
|
3
|
2
|
+1
|
2
|
2
|
0
|
5
|
6
|
-1
|
3
|
2
|
+1
|
|
Contributor
|
N. Sobhkhiz
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Contributor
|
Roya Zoraghi
|
2
|
2
|
0
|
4
|
5
|
-1
|
3
|
2
|
+1
|
4
|
5
|
-1
|
2
|
2
|
0
|
2
|
1
|
+1
|
|
Contributor
|
Ali Haeri Rouhani
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Contributor
|
Yvonne Bone
|
49
|
47
|
+2
|
7
|
26
|
-19
|
201
|
201
|
0
|
30
|
39
|
-9
|
50
|
47
|
+3
|
201
|
201
|
0
|
R1: The first retrieval rank
R2: The second retrieval rank
R3: The difference achieved
Appendix C: The rankings of the keywords in
the control group
|
|
|
AlltheWeb
|
AltaVista
|
Excite
|
Google
|
Lycos
|
WebCrawler
|
|
Element
|
Keyword
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
R1
|
R2
|
R3
|
|
Title
|
Boolean Algebras
|
201
|
201
|
0
|
74
|
76
|
-2
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Join Spaces
|
1
|
1
|
0
|
5
|
5
|
0
|
201
|
8
|
+193
|
16
|
12
|
+4
|
2
|
1
|
+1
|
201
|
8
|
+193
|
|
Title
|
Infinitely
Divisible Distributions
|
28
|
21
|
+7
|
32
|
35
|
-3
|
201
|
38
|
+163
|
32
|
28
|
+4
|
28
|
21
|
+7
|
201
|
30
|
+171
|
|
Title
|
Cytogenetic
Biomonitoring
|
25
|
36
|
-11
|
2
|
5
|
-3
|
201
|
201
|
0
|
12
|
21
|
-9
|
25
|
32
|
-7
|
201
|
201
|
0
|
|
Title
|
Aromatic solvents
|
201
|
201
|
0
|
22
|
20
|
+2
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Marine
Palynomorph
|
22
|
20
|
+2
|
2
|
1
|
+1
|
25
|
7
|
+18
|
3
|
4
|
-1
|
22
|
17
|
+5
|
25
|
7
|
+18
|
|
Title
|
Sequence
boundaries
|
201
|
201
|
0
|
40
|
38
|
+2
|
201
|
201
|
0
|
129
|
82
|
+47
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Upper Jurassic
Sediments
|
6
|
56
|
-50
|
3
|
201
|
-198
|
201
|
201
|
0
|
18
|
16
|
+2
|
6
|
52
|
-46
|
201
|
201
|
0
|
|
Title
|
Northern
Switzerland
|
107
|
201
|
-94
|
201
|
201
|
0
|
201
|
201
|
0
|
46
|
30
|
+16
|
109
|
201
|
-92
|
201
|
24
|
+177
|
|
Title
|
Climate
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Flora
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Title
|
Oil Field Areas
|
81
|
7
|
+74
|
1
|
1
|
0
|
201
|
29
|
+172
|
28
|
10
|
+18
|
82
|
7
|
+75
|
201
|
29
|
+172
|
|
Title
|
South-West of Iran
|
66
|
73
|
-7
|
9
|
8
|
+1
|
201
|
37
|
+164
|
201
|
42
|
+159
|
66
|
73
|
-7
|
201
|
38
|
+163
|
|
Subject
|
Hypergroup
|
30
|
48
|
-18
|
10
|
8
|
+2
|
201
|
201
|
0
|
21
|
19
|
+2
|
32
|
48
|
-16
|
201
|
201
|
0
|
|
Subject
|
Algebraic
hyperstructure
|
1
|
1
|
0
|
1
|
1
|
0
|
2
|
1
|
+1
|
1
|
1
|
0
|
1
|
1
|
0
|
2
|
1
|
+1
|
|
Subject
|
Infinite
divisibility
|
201
|
201
|
0
|
70
|
75
|
-5
|
201
|
201
|
0
|
42
|
55
|
-13
|
201
|
201
|
0
|
1
|
201
|
-200
|
|
Subject
|
Strictly stable
distributions
|
2
|
2
|
0
|
1
|
1
|
0
|
201
|
2
|
+199
|
3
|
4
|
-1
|
2
|
2
|
0
|
201
|
2
|
+199
|
|
Subject
|
Cauchy
distribution
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Normal
distribution
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Characteristics
function
|
201
|
201
|
0
|
105
|
120
|
-15
|
201
|
201
|
0
|
199
|
123
|
+76
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Unimodality
|
201
|
201
|
0
|
48
|
79
|
-31
|
201
|
201
|
0
|
125
|
146
|
-21
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Chromosomal
Aberrations
|
201
|
201
|
0
|
183
|
201
|
-18
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Lymphocytes
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Occupational
Exposure
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Organic Solvents
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Dinoflagellates
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Rhodano-Swabian
basin
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Subject
|
Floristic
composition
|
201
|
201
|
0
|
119
|
123
|
-4
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
Saharo-Sindian
region
|
3
|
4
|
-1
|
1
|
1
|
0
|
201
|
1
|
+200
|
2
|
1
|
+1
|
3
|
2
|
+1
|
201
|
1
|
+200
|
|
Subject
|
Plant geography
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
201
|
201
|
0
|
|
Subject
|
SW. Iran
|
47
|
53
|
-6
|
17
|
15
|
+2
|
201
|
50
|
+151
|
66
|
47
|
+19
|
47
|
51
|
-4
|
201
|
48
|
+153
|
|
Subject
|
Khuzistan
|
23
|
22
|
+1
|
21
|
25
|
-4
|
201
|
201
|
0
|
34
|
40
|
-6
|
23
|
22
|
+1
|
201
|
201
|
0
|
|
Creator
|
Ali Reza Ashrafi
|
5
|
6
|
-1
|
7
|
11
|
-4
|
7
|
16
|
-9
|
10
|
11
|
-1
|
5
|
6
|
-1
|
7
|
16
|
-9
|
|
Creator
|
M. Hossein
Alamatsaz
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Creator
|
Hossein Mozdarani
|
6
|
9
|
-3
|
5
|
5
|
0
|
4
|
5
|
-1
|
8
|
6
|
+2
|
6
|
9
|
-3
|
4
|
5
|
-1
|
|
Creator
|
Ebrahim
Ghasemi-Nejad
|
7
|
9
|
-3
|
3
|
2
|
+1
|
7
|
5
|
+2
|
2
|
2
|
0
|
7
|
8
|
-1
|
7
|
5
|
+2
|
|
Creator
|
Ebrahim Alaie
|
1
|
1
|
0
|
1
|
2
|
-1
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Creator
|
Shirazeh Arghami
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
1
|
1
|
0
|
|
Creator
|
A. Ghahreman
|
1
|
2
|
-1
|
3
|
3
|
0
|
12
|
2
|
+10
|
3
|
3
|
0
|
1
|
2
|
-1
|
12
|
2
|
+10
|
R1: The first retrieval rank
R2: The second retrieval rank
R3: The difference achieved
|