In this issue

Abstract

The editor of this special topics issue has covered those papers and this column will address only the three regular submissions included in this issue; two on citation behavior and one on the efficacy of query expansion. In the first of these, Brown uses 197 e-prints from Elsevier's Chemistry Preprint Server, CPS, for citation analysis to determine if they are cited in the chemical literature, as well as surveys of the 116 CPS authors and leading chemical journal editors ( as shown in Journal Citation Reports) to determine the use and influence of the e-print in chemistry. The highest viewed and discussed e-prints in July of 2001 were checked for appearance as printed papers in December of 2001. Of these, 32% later appeared in the regular journal literature. The 28% of editors responding, with one exception, did not accept papers in the CPS or had no policy. On the acceptance of citations to e-prints, 27% said yes, 36% no, and 36% chose the other category. The 60 responding authors (52%) reported the submission of 1.7 papers to CPS, visit the site regularly, and 78% indicated they also contribute to peer reviewed journals. No citations to e-prints were found in the Web of Science from 2000 to 2001. The only citations to CPS documents from CPS documents were two self citations. While CPS is utilized and valued by its users, it is not integrated into the web of chemistry citation.Burrell predicts the future citation behavior for sources in an homogeneous set of papers each having a given number of current citations with the thought that the estimate of a paper's total future citations is an indicator of its impact. Each paper is assumed to have a latent rate at which citations may be acquired which constitutes a random variable over the collection. Obsolescence is represented by a cumulative distribution function on time, and the number of total citations at a given time as a Poisson distribution, the mean of which increases with time to a finite limit. The distribution of the additional number of citations to a paper during an interval of time can then be found permitting the computation of an expected total number of citations. The latent rate distribution may be adequately modeled by the gamma distribution, leading to a negative binomial distribution for the conditional distribution of citations, and a proof that the mean number of additional citations is a linear function of the current number of citations, which expresses the success breeds success principle. It is also possible to find the probability that an un-cited paper will be cited in a given time interval, or that a paper having received a given number of citations will receive no more in such an interval. It is not totally clear how to estimate the gamma distribution parameters, and there is currently no empirical evidence to support the model.Finally, using TREC-9 Query Track performance data, Alemayehu ranks the query expansion methods by mean average precision. Three INQUERY, three SMART, and one OKAPI runs are tested to determine if significant v...