B U L L E T I N
In this issue we begin a new feature in which we will publish structured, “bottom-line” abstracts of selected articles from the Journal of the American Society for Information Science and Technology (JASIST) to improve dissemination of research findings that might be of general interest. The fact that an article does not appear here certainly does not mean that it is of no interest to practitioners. First, there was a start date when JASIST began notifying authors of accepted articles of the opportunity to submit abstracts to the Bulletin, and no attempt was made to solicit retrospectively. Some articles may, therefore, have been accepted before we initiated this project. Second, submission is optional, and third the Editor can only choose a few due to space restrictions. We would appreciate your comments and input to Bulletin@asis.org.
From JASIST, v. 53 (13)
Slone, D. (2002). The influence of mental models and goals on user search patterns during web interaction, pp. 1152-1169.
Study and Results: This study used observations and interviews to determine the influence of situational goals and mental models on Web searching behavior. Users with educational or job-related situational goals were found to be more motivated than users with recreational or personal goals. Participants seeking educational or job-related information demonstrated mature mental models by utilizing many search approaches and accessing a high number of pages. In contrast, users seeking information for recreational or personal use utilized few search approaches and accessed few Web pages. When unmotivated users with immature mental models encountered problems, they abandoned the Internet as an information source, while highly motivated users persisted longer whether or not their mental models were mature.
What's New? Unlike other studies of end-user searching, this inquiry uses participant generated queries, a public library setting and participants from various age groups, ethnic groups and skill levels. Additionally, it compares Web online catalog searching to other types of Web searching.
Limitations: The study is limited to 30 public library users. It is a snapshot of activity during a specific time period. Thus, results may vary depending on Web browsers, links, time of day, type of library and other factors.
Aljlayl, M., Frieder, O. & Grossman, D. (2002). On bi-directional English-Arabic search, pp.1139-1151.
Study and Results: We demonstrate that language translation ambiguities within a query can be addressed without complex natural language processing. To show this we took a set of cross-lingual queries and evaluated their retrieval accuracy using several machine-readable dictionary-translation algorithms. The idea is to take an English query, translate it into Arabic and then run it against an Arabic document collection. Our algorithms use a machine-readable dictionary to translate and resolve ambiguities by retaining only those query terms whose translation from English to Arabic and then back to English returns only the original query term. This approach, referred to as the two-phase method, yields statistically better results than the prior state-of-the-art approaches on the TREC collection.
What’s New?: We describe a variety of new algorithms for machine-readable translation of queries. Work in cross-lingual information retrieval of Arabic is relatively new, as the TREC Arabic collection has only recently been made available for research. Our results are among the first published for this collection.
Limitations: As with any tests on a single test collection and a single set of queries, it is always the case that results may vary as collections and queries change. With continued use of a variety of Arabic collections and queries, we hope to alleviate this concern in the future.
Abt, H. & Garfield, E. (2002). Is the relationship between numbers of references and paper lengths the same for all sciences? pp. 1106-1112.
Study and Results: The use of citation counts to evaluate research papers is based upon the assumption that authors in various fields of science use references in similar ways. If authors in one field use references sparingly and those in another field place many references in their papers, the citation counts in those different fields will be very different and may not reflect the relative merits of their papers.
The authors explore the use of references in 41 journals in the physical, life and social sciences. We surveyed about 200 papers per journal and found that the numbers of references per paper depended primarily upon the paper lengths and only very slightly upon journal impact factors and decade of publication. In each journal there is a linear relation between numbers of references and paper length; that relation is the same for all the journals in each field of science.
The papers in the social sciences average longer (about 15 1000-word pages) than those in general medicine (5 pages), immunology (8 pages), biochemistry (9 pages) and the physical sciences (11 pages). The surprising result is that the mean number of references is nearly the same (41 7) for all those sciences. The error is easily attributed to format differences, such as whether the references quote titles of journal papers.
What's New?: We conclude that it is valid to compare citation counts for papers in different sciences because they average the same number of references per paper.
Limitations: Review papers have twice as many references as original research papers. Data-rich papers have fewer references.
From JASIST, v. 53 (14)
Editor’s Note: Most of JASIST 53 (14) is a Perspectives section on “Methods for Studying Information Seeking and Use,” which I highly recommend. I do not have abstracts for them.
Whitley, K. M. (2002). Analysis of SciFinder Scholar and Web of Science citation searches, pp. 1210-1215.
Study and Results: University administrations are fond of citation counts as an evaluation criterion for faculty tenure and promotion, particularly in the sciences. This paper analyzes the duplication and uniqueness of sample citation searches for chemistry researchers in the SciFinder Scholar (Chemical Abstracts) and Web of Science (Science Citation Index) databases. Findings show approximately a 70 percent overlap in citation coverage between SFS and WOS, which means that the average researcher in this study would miss almost a third of the possible citations by relying on only one database. Also, each database misses researchers’ articles in major journals. The most complete search requires using both databases.
What’s New?: Comparison of citation searching in SciFinder Scholar and Web of Science shows that relying on either index alone does not produce complete citation totals for individual scholars.
Limitations: The sample contains only 30 researchers from different chemistry sub-disciplines. Interesting future research might focus on more detailed analysis of characteristics of the sub-disciplines.
Copyright © 2003, American Society for Information Science and Technology