Enhancing citation recommendation using citation network embedding

Scientometrics - Automatic recommendation of citations has been a focal point of research in scholarly digital libraries. Many graph-based citation recommendation algorithms have been proposed;...     

A scientific citation recommendation model integrating network and text representations

Scientometrics - The number of scientific papers is increasing in the rapid growth. How to make paper acquisition efficient and provide effective citation recommendation is essential for...     

Task recommendation in crowdsourcing systems: A bibliometric analysis

Existing studies on task recommendation in crowdsourcing systems provide additional insights into the field from their perspectives, methodologies, frameworks, and disciplines, resulting in a highly productive but unorganized knowledge domain. This paper is motivated to exploit bibliometric techniques to derive insights that exceed the boundaries of individual systems and identify the potentially transformative changes from 268 published articles. The explicit features (i.e., affiliation, author, citation, and keywords) and implicit information (i.e., topic distribution, potential structure, hidden insights, and evolutionary trend) of domain literature are discovered by network analysis, cluster analysis, and timeline analysis. We summarize the generic framework based on knowledge domain structure and highlight the position of knowledge source, especially textual information, in task recommendation models. Drawing on the Shneider four-stage model, the temporal evolution trend is graphically illustrated to emphasize avenues for future research. Our study conveys accumulated and synthesized specialty knowledge to researchers or newcomers to help them design, initiate, implement, manage, and evaluate recommender systems in crowdsourcing.     

Does the higher citation of collaborative research differ from region to region? A case study of Economics

Many studies have found that collaborative research is, in general, more highly cited than non-collaborative research. This paper describes an investigation into the extent to which the association between high citation and collaboration for Economics articles published in 2000 varies from region to region and depends on the choice of indicator of citation level. Using data from the Social Science Citation Index (SSCI) for 18 countries, 17 American states and four indicators of citation level the citation levels of the collaborative articles are compared with the citation levels of the non-collaborative articles. The main findings are that: (a) for every country and every indicator the mean citation level of the collaborative articles was at least as high as that for the non-collaborative articles, but for five US states and for at least one other indicator the citation level of collaborative articles was lower than that of non-collaborative articles, and (b) the extent to which collaborative articles were more highly cited varied considerably from country to country, from state to state, and from indicator to indicator. This indicates the importance of using multiple indicators when investigating citation advantage since the choice of indicator can change the results.     

Going much beyond the Durfee square: enhancing the <Emphasis Type="Italic">h</Emphasis><Subscript>T</Subscript> index

The objective of this research is to develop a new patent bibliometric performance measure by using modified citation rate analyses with dynamic backward citation windows. Cited half-life employed in bibliometrics was adopted in order to establish a model of annual patent backward citation windows. Based on the dynamic behavior of backward citation windows, the annual backward patent citation rates for each technology domain can be calculated to measure its bibliometric performance. It was found that the dynamic backward citation window represents more accurately the citation cycle time which is a key factor on technology assessment. Because different technology domain may have disparate attributes, a normalized backward citation rate was developed to measure the corresponding rank for each domain respect to the entire industry. Three technology domains were then chosen for demonstrative case studies which represent semiconductor, LCD, and drug industries.     

Self-citations at the meso and individual levels: effects of different calculation methods

This paper focuses on the study of self-citations at the meso and micro (individual) levels, on the basis of an analysis of the production (1994–2004) of individual researchers working at the Spanish CSIC in the areas of Biology and Biomedicine and Material Sciences. Two different types of self-citations are described: author self-citations (citations received from the author him/herself) and co-author self-citations (citations received from the researchers’ co-authors but without his/her participation). Self-citations do not play a decisive role in the high citation scores of documents either at the individual or at the meso level, which are mainly due to external citations. At micro-level, the percentage of self-citations does not change by professional rank or age, but differences in the relative weight of author and co-author self-citations have been found. The percentage of co-author self-citations tends to decrease with age and professional rank while the percentage of author self-citations shows the opposite trend. Suppressing author self-citations from citation counts to prevent overblown self-citation practices may result in a higher reduction of citation numbers of old scientists and, particularly, of those in the highest categories. Author and co-author self-citations provide valuable information on the scientific communication process, but external citations are the most relevant for evaluative purposes. As a final recommendation, studies considering self-citations at the individual level should make clear whether author or total self-citations are used as these can affect researchers differently.     

Exploiting Structural Similarities to Classify Citations

Citations play an important role in the scientific community by assisting in measuring multifarious policies like the impact of journals, researchers, institutions, and countries. Authors cite papers for different reasons, such as extending previous work, comparing their study with the state-of-the-art, providing background of the field, etc. In recent years, researchers have tried to conceptualize all citations into two broad categories, important and incidental. Such a categorization is very important to enhance scientific output in multiple ways, for instance, (1) Helping a researcher in identifying meaningful citations from a list of 100 to 1000 citations (2) Enhancing the impact factor calculation mechanism by more strongly weighting important citations, and (3) Improving researcher, institutional, and university rankings by only considering important citations. All of these uses depend upon correctly identifying the important citations from the list of all citations in a paper. To date, researchers have utilized many features to classify citations into these broad categories: cue phrases, in-text citation counts, and metadata features, etc. However, contemporary approaches are based on identification of in-text citation counts, mapping sections onto the Introduction, Methods, Results, and Discussion (IMRAD) structure, identifying cue phrases, etc. Identifying such features accurately is a challenging task and is normally conducted manually, with the accuracy of citation classification demonstrated in terms of these manually extracted features. This research proposes to examine the content of the cited and citing pair to identify important citing papers for each cited paper. This content similarity approach was adopted from research paper recommendation approaches. Furthermore, a novel section-based content similarity approach is also proposed. The results show that solely using the abstract of the cited and citing papers can achieve similar accuracy as the state-of-the-art approaches. This makes the proposed approach a viable technique that does not depend on manual identification of complex features.     

Citation time window choice for research impact evaluation

This paper aims to inform choice of citation time window for research evaluation, by answering three questions: (1) How accurate is it to use citation counts in short time windows to approximate total citations? (2) How does citation ageing vary by research fields, document types, publication months, and total citations? (3) Can field normalization improve the accuracy of using short citation time windows? We investigate the 31-year life time non-self-citation processes of all Thomson Reuters Web of Science journal papers published in 1980. The correlation between non-self-citation counts in each time window and total non-self-citations in all 31 years is calculated, and it is lower for more highly cited papers than less highly cited ones. There are significant differences in citation ageing between different research fields, document types, total citation counts, and publication months. However, the within group differences are more striking; many papers in the slowest ageing field may still age faster than many papers in the fastest ageing field. Furthermore, field normalization cannot improve the accuracy of using short citation time windows. Implications and recommendations for choosing adequate citation time windows are discussed.     

Self-rated driving and driving safety in older adults

Many U.S. states rely on older adults to self-regulate their driving and determine when driving is no longer a safe option. However, the relationship of older adults’ self-rated driving in terms of actual driving competency outcomes is unclear. The current study investigates self-rated driving in terms of (1) systematic differences between older adults with high (good/excellent) versus low (poor/fair/average) self-ratings, and (2) the predictive nature of self-rated driving to adverse driving outcomes in older adults (n = 350; mean age 73.9, SD = 5.25, range 65–91). Adverse driving outcomes included self-reported incidences of (1) being pulled over by the police, (2) receiving a citation, (3) receiving a recommendation to cease or limit driving, (4) crashes, and (5) state-reported crashes. Results found that older drivers with low self-ratings reported more medical conditions, less driving frequency, and had been given more suggestions to stop/limit their driving; there were no other significant differences between low and high self-raters. Logistic regression revealed older drivers were more likely to have a state-reported crash and receive a suggestion to stop or limit driving. Men were more likely to report all adverse driving outcomes except for receiving a suggestion to stop or limit driving. Regarding self-rated driving, older adults with high ratings were 66% less likely (OR = 0.34, 95% CI = 0.14–0.85) to have received suggestions to limit or stop driving after accounting for demographics, health and driving frequency. Self-ratings were not predictive of other driving outcomes (being pulled over by the police, receiving a citation, self-reported crashes, or state-reported crashes, ps > 0.05). Most older drivers (85.14%) rated themselves as either good or excellent drivers regardless of their actual previous citation or crash rates. Self-rated driving is likely not related to actual driving proficiency as indicated by previous crash involvement in older adults. Suggestions from other individuals to limit or cease driving may be more influential on self-ratings.     

Exploring dynamic research interest and academic influence for scientific collaborator recommendation

In many cases, it is time-consuming for researchers to find proper collaborators who can provide researching guidance besides simply collaborating. The Most Beneficial Collaborators (MBCs), who have high academic level and relevant research topics, can genuinely help researchers to enrich their research. However, how can we find the MBCs? In this paper, we propose a most Beneficial Collaborator Recommendation model called BCR. BCR learns on researchers’ publications and associates three academic features: topic distribution of research interest, interest variation with time and researchers’ impact in collaborators network. We run a topic model on researchers’ publications in each year for topic clustering. The generated topic distribution matrix is fixed by a time function to fit the interest dynamic transformation. The academic social impact is also considered to mine the most prolific researchers. Finally, a TopN MBCs recommendation list is generated according to the computed score. Extensive experiments on a dataset with citation network demonstrate that, in comparison to relevant baseline approaches, our BCR performs better in terms of precision, recall, F1 score and the recommendation quality.     

A patent-journal citation network on prostaglandins

The paper describes work on a citation network starting from ten U.S. patents and comprising 1204 journal articles and 43 patents in the field of prostaglandins. A journal citation network was created independently from 85 recent articles on the subject. It was found that the two networks were virtually identical except that the former included patents and was therefore more complete. It was found that the most highly cited references were those widely regarded as being key turning points in the development of the subject.The results demonstrate that in a subject field which incorporates patentable subject-matter matter, a patent/journal citation network is fuller and more reliable than a patent citation network or a journal citation network on their own.     

Citation ranking versus expert judgment in evaluating communication scholars: Effects of research specialty size and individual prominence

Numerous attempts have been made to validate the use of citation as an evaluation method by comparing it with peer review. Unlike past studies using journals, research articles or universities as the subject matter, the present study extends the comparison to the ranking of individual scholars. Results show that citation ranking and expert judgment of communication scholars are highly correlated. The citation method and the expert judgment method are found to work better in smaller research areas and yield more valid evaluation results for more prominent scholars.     

Citation analysis with microsoft academic

In this article, we compare publication and citation coverage of the new Microsoft Academic with all other major sources for bibliometric data: Google Scholar, Scopus, and the Web of Science, using a sample of 145 academics in five broad disciplinary areas: Life Sciences, Sciences, Engineering, Social Sciences, and Humanities. When using the more conservative linked citation counts for Microsoft Academic, this data-source provides higher citation counts than both Scopus and the Web of Science for Engineering, the Social Sciences, and the Humanities, whereas citation counts for the Life Sciences and the Sciences are fairly similar across these three databases. Google Scholar still reports the highest citation counts for all disciplines. When using the more liberal estimated citation counts for Microsoft Academic, its average citations counts are higher than both Scopus and the Web of Science for all disciplines. For the Life Sciences, Microsoft Academic estimated citation counts are higher even than Google Scholar counts, whereas for the Sciences they are almost identical. For Engineering, Microsoft Academic estimated citation counts are 14% lower than Google Scholar citation counts, whereas for the Social Sciences this is 23%. Only for the Humanities are they substantially (69%) lower than Google Scholar citations counts. Overall, this first large-scale comparative study suggests that the new incarnation of Microsoft Academic presents us with an excellent alternative for citation analysis. We therefore conclude that the Microsoft Academic Phoenix is undeniably growing wings; it might be ready to fly off and start its adult life in the field of research evaluation soon.

     

The money value of citations to single-authored and multiple-authored articles

This note presents evidence for the surprising conclusion that a citation to a multiple-authored article is worth more to its author than a citation to a single-authored article.     

International and domestic co-publishing and their citation impact in different disciplines

This paper studies disciplinary differences in citation impacts of different types of co-publishing. The citation impacts of international, domestic inter-organizational and domestic intra-organizational co-publications, and single-authored publications, are compared. In particular, we examine the extent to which the number of authors explains the potential differences in citation impacts when compared to the influence of different types of international and domestic collaborations. The analysis is based on Finland’s publications in Thomson Reuters Web of Science database in 1990–2008. Finland is a small country, thus, it has fewer opportunities to find collaborators inside own country when compared to larger countries. Finland’s science policy has underlined internationalization and research collaboration as key means to increase the quality and impact of Finnish research. This study indicates that both international and domestic co-publishing have steadily increased during the past two decades in all disciplinary groups. International co-publications gain on average more citations than domestic co-publications. In natural sciences and engineering, co-authorship explains only a small proportion of variability in publications’ citation rates. When the effect of the number of authors is taken into account there are no big differences in citation impacts between international and domestic co-publications. However, international co-publications by ten authors or more gather significantly more citations than other publications. In humanities, the difference in citation impacts between co-authored publications in relation to single-authored publications is significant. However, international co-publications are not on average more highly cited in relation to domestic co-publications in humanities.     

Patterns of annual citation of highly cited articles and the prediction of their citation ranking: A comparison across subjects

High citation is associated with research quality and consequently findings on highly cited articles are useful to increase understanding of the factors that produce high quality research. This study explores highly cited articles in six subjects, focusing on late citation and peak citation years. Longitudinal citation patterns were found to be highly varied and, on average, different from the remaining articles in each subject. For four of the six subjects, there is a correlation of over 0.42 between the percentage of early citations and total citation ranking but more highly ranked articles had a lower percentage of early citations. Surprisingly, for highly cited articles in all six subjects the prediction of citation ranking of from the sum of citations during their first six years was less accurate than prediction using the sum of the citations for only the fifth and sixth year.     

Citation success over time: theory or empirics?

This study investigates the citation patterns of theoretical and empirical papers published in a top economics journal, namely American Economic Review, over a period of almost 30 years, while also exploring the determinants of citation success. The results indicate that empirical papers attract more citation success than theoretical studies. However, the pattern over time is very similar. Moreover, among empirical papers it appears that the cross-country studies are more successful than single country studies focusing on North America data or other regions.     

A study on the citation situation within the citing paper: citation distribution of references according to mention frequency

Citation impact indicators play a significant role in evaluating the scientific research activity. Most of citation impact indicators are based on the citation count that the publication is cited as a reference in the other publications, but the difference between each citation situation was not considered. Normally, the number of citations that a publication is cited in the other publications may represent the formal quality of the publication. Similarly, the number of times that a publication is really mentioned within the citing publication, it may also represent the formal quality of the citation. We have examined about how many times each reference was really mentioned within the citing publications and studied about the citation situation within the citing publications. We verified that the citation distribution of references according to the mention frequency follows the Generalized Pareto distribution. The results showed that about 20% of total references were mentioned three and more times, and the number of citation mentions for the about 50% of total references were from about 20% of the total references in the citing publications.     

Can scientific impact be judged prospectively? A bibliometric test of Simonton"s model of creative productivity

Simonton"s (1997) model of creative productivity, based on a blind variation-selection process, predicts scientific impact can only be evaluated retrospectively, after recognition has been achieved. We test this hypothesis using bibliometric data from the Human Factors journal, which gives an award for the best paper published each year. If Simonton"s model is correct, award winning papers would not be cited much more frequently than non-award winning papers, showing that scientific success cannot be judged prospectively. The results generally confirm Simonton"s model. Receipt of the award increases the citation rate of articles, but accounts for only 0.8% to 1.2% of the variance in the citation rate. Consistent with Simonton"s model, the influence of the award on citation rate may reflect a selection process of an elite group of reviewers who are representative of the larger peer group that eventually determines the citation rate of the article. Consistent with Simonton"s model, author productivity accounts for far more variance in the authors" total citation rate (58.9%) and in the citation rate of the authors" most cited article (12.6%) than does award receipt. This revised version was published online in August 2006 with corrections to the Cover Date.     

Is citation analysis a legitimate evaluation tool?

A comprehensive discussion on the use of citation analysis to rate scientific performance and the controversy surrounding it. The general adverse criticism that citation counts include an excessive number of negative citations (citations to incorrect results worthy of attack), self-citations (citations to the works of the citing authors), and citations to methodological papers is analyzed. Included are a discussion of measurement problems such as counting citations for multiauthored papers, distinguishing between more than one person with the same last name (homographs), and what it is that citation analysis actually measures. It is concluded that as the scientific enterprise becomes larger and more complex, and its role in society more critical, it will become more difficult, expensive and necessary to evaluate and identify the largest contributors. When properly used, citation analysis can introduce a useful measure of objectivity into the evaluation process at relatively low financial cost.Modification of a chapter in E. Garfield:Citation Indexing: Its Theory and Application in Science, Technology and the Humanities, New York, Wiley, 1979.