Probabilistic Datalog: Implementing Logical Information Retrieval for Advanced Applications
Norbert Fuhr To begin this issue, Fuhr demonstrates that a probabilistic implementation of the query
language Datalog can be used as a retrieval language. The high expressiveness of this predicate logic language allows the use of vague queries on imprecise data. All entities yielding a non-zero probability must be
processed which leads to a high use of resources. No evaluation of effectiveness compared to other models is provided. Interface Metaphors and Logical Analogues: A Question of Terminology
Anne Hamilton It
is possible to question metaphor as a design principle, since varied interpretation is common. Hamilton interviews fifteen Macintosh users, practicing academics in a university department, on their Macintosh learning
experiences, and subjects the resulting video recordings to content analysis. The desktop metaphor was generally found to be vague, unclear and unimportant. The typewriter metaphor was confusing to younger subjects who
had not used typewriters, but easily understood by others. There were mismatches and most who thought the metaphor helpful found it so only for a learning period. The trash can metaphor remained generally relevant and
current but mismatches like disc ejection by moving the disc icon to the trash can icon evoked strong negative reaction in the subjects even after years of use. It appears users expect logical analogs, not metaphors.
Citation Ranking Versus Peer Evaluation of Senior Faculty Research Performance: A Case Study of Kurdish Scholarship
Lokman I. Meho and Diane H. Sonnenwald Five senior scholars in Kurdish studies were ordered
by Meho and Sonnenwald as to their peer rankings, number of citations since their dissertations, quality of citations, and quality of books published, where quality was determined by content analysis of the items
produced. Each participant contributed a bibliography on his Kurd production which was then subjected to a multiple database confirmation before citations were collected. Book reviews were searched and evaluative
comments classed. Negative citations were minimal in number. All four measures result in the same rankings, except for a second third place switch in peer rankings and a forth third place switch in adjusted citation
count. Citation, citation content and peer review have a significant rank order correlation. Not so for reviews. All measures give similar results for top and bottom ranked scholars.
Publication Trends of Doctoral Students in Three Fields from 1965-1995
Wade M. Lee Using three disciplines from the Dissertation Abstracts index in every fifth year from 1965 to 1995, Lee selected dissertations
randomly in sufficient numbers to give a 95% confidence level. The publication record of each author was then searched for the dissertation year, and three years before and one year after that date in appropriate
disciplinary indexes. Chemists generally publish more as doctoral students than do psychologists, whose output exceeds literature scholars. Rates are lower than published rates for the professions, but the publication
rate of chemists is increasing over time. The appearance of single author papers declines over time. Methods for Accrediting Publications to Authors or Countries: Consequences for Evaluation Studies
Leo Egghe, Ronald Rousseau, and Guido Van Hooydonk The Influence of Publication Delays on the Observed Aging Distribution of Scientific Literature
Leo Egghe and Ronald Rousseau Egghe and Rousseau find
that the variation in delay in publication of papers citing a given paper will affect evaluation of aging, defined as declining use as indicated by citations. The observed distribution is a result of convolution of the
aging distribution and the publication delay distribution. The larger the average publication delay, the larger the shift in the average of the aging curve. With two Poisson distributions the observed distribution
values are higher than either the delay distribution or the undisturbed aging distribution. Assuming two exponential distributions, which often approximate large values, the convolution of the two decreasing curves will
yield a unimodal distribution and the initially dominant distribution will have the least influence for these values. With two lognormal distributions the resulting integral is unsolvable using elementary functions but
the convolution can be demonstrated to be unimodal. The resulting distributions all indicate that the observed aging distribution should be convex, but observed data seem to result in concave distributions.
Semantic Similarities Between a Keyword Database and a Controlled Vocabulary Database: An Investigation in the Antibiotic Resistance Literature
Jian Qin Qin compares SCI keyword plus indexing to MEDLINE
indexing on 414 documents on antibiotic resistance found in both databases. Terms were coded to indicate whether or not they were in common in matching records, and whether hierarchically different, or synonymic forms
of commonality were present, as well as to give different spellings and word forms a common textual code. Using the inclusion index, a variation of the Jaccard similarity measure for use with other than symmetric data,
she finds 41% of terms in SCI match MEDLINE and 33% of terms in MEDLINE match SCI. Large within document dissimilarities in indexing exist. About 10% of document indices resembled one another and 25% exhibited partial
resemblance. Comparisons of the top twenty terms in each vocabulary show some meaningful similarities reflecting the literature topic. Readers, Authors, and Page Structure: A Discussion of Four Questions Arising
from a Content Analysis of Web Pages
Stephanie W. Haas and Erika S. Grams Haas and Grams collect 331 randomly selected Web pages with links. The first generation target pages were then selected for each source,
and URL, title, header, and a description based upon researcher-generated classes, stored for each source and target page. Source page anchors were described by text, image description, setting, and by the apparent
reason for their existence. Seven page types were identified, but while many pages were clearly of one type, many were also combinations of several types. Ninety percent of isolated anchors are noun structures of some
sort, and more than two thirds are navigational in nature. Eighty six percent of all links are within the same site. Application of Dublin Core Metadata in the Description of Digital Primary Sources in Elementary
School Classrooms
Anne J. Gilliland-Swetland, Yasmin B. Kafai, and William E. Landis Gilliland-Swetland, Kafai, and Landis examine the work of fifty-eight students in 4th and 5th grade classes working in teams
using Dublin Core metadata elements, assigned to selected materials from a naturalist's work to develop descriptions from their own field observation notebooks of their photographs of the same wetland sites. Source
description of place was inaccurate in three cases. The author description always contained accurate photographer information. Accuracy of recording relationships to other pictures was at 59% with nearly 40% left blank,
and the subject categories chosen were 96% accurate. Titles were never judged inaccurate but sometimes on an incorrect. hierarchic level. The description of the picture was judged sometimes inaccurate, and sometimes
less or more detailed than the title. |
