Source-linked AI summary
A Bibliometric Model for Identifying Emerging Research Topics
Qi Wang
TL;DR
Identifying emerging research topics is difficult because definitions are ambiguous and bibliometric indicators are not always clearly linked to the concepts they measure. The paper adapts a definition, proposes corresponding identification criteria, and identifies several topics, presenting its approach as conceptually straightforward and operationally transparent.
Problem
Emergence definitions are ambiguous, and prior bibliometric indicators are not always clearly connected to the concepts they operationalize.
Source: Introduction · p. 2, Further Discussion on Related Work · p. 5Method
The study adapts Rotolo et al. (2015)’s definition for research topics and develops identification criteria linked to the attributes of emergence.
Source: Further Discussion on Related Work · p. 5, Definition of Emerging Research Topics · p. 5Results
Using two parameter sets, the study identified several emerging topics and evaluated the approach through demonstrations, prior studies, and an in-depth LIS analysis.
Source: Discussion and Conclusions · p. 19Takeaways & Limitations
The approach is presented as conceptually straightforward and operationally transparent because it connects emergence attributes with indicators.
Source: Discussion and Conclusions · p. 19The method uses WoS articles and reviews, so it may be less useful in fields where monographs or conference papers carry novel knowledge.
Source: Discussion and Conclusions · p. 19
Abstract
from arXiv · showhide
Detecting emerging research topics is essential, not only for research agencies but also for individual researchers. Previous studies have created various bibliographic indicators for the identification of emerging research topics. However, as indicated by Rotolo et al. (2015), the most serious problems are the lack of an acknowledged definition of emergence and incomplete elaboration of the linkages between the definitions that are used and the indicators that are created. With these issues in mind, this study first adjusts the definition of an emerging technology that Rotolo et al. (2015) have proposed in order to accommodate the analysis. Next, a set of criteria for the identification of emerging topics is proposed according to the adjusted definition and attributes of emergence. By the use of two sets of parameter values, several emerging research topics are identified. Finally, evaluation tests are conducted by demonstration of the proposed approach and comparison with previous studies. The strength of the present methodology lies in the fact that it is fully transparent, straightforward, and flexible.
Introduction
The section distinguishes identifying emerging topics from measuring topics already identified, while noting ambiguity in emergence definitions and weak links to bibliometric indicators. It reviews approaches to defining research topics and motivates a transparent method connecting concepts to indicators.
Introduction
- Prior quantitative studies often measured pre-identified topics, while funding agencies and policy makers need methods that discover emerging topics.Only a few studies had proposed methodologies aimed at identification rather than measurement.Source: Introduction · pp. 1–2
- Definitions and attributes of emergence remain ambiguous, and bibliometric indicators are often weakly connected to the concepts they are meant to operationalize.The authors identify this gap as a general limitation of studies that seek to identify emerging topics.Source: Introduction · p. 2
Related Work
- Prior quantitative bibliometric studies commonly build publication clusters from direct citation relations, then apply criteria to identify emerging topics.Research-topic construction also uses text-based, other citation-based, hybrid, and topic-modeling approaches.Source: Related Work · p. 3, Establishing Research Topics · p. 3
Establishing Research Topics
- Because journal classifications such as WoS and Scopus are coarse-grained, most identification studies establish finer-grained research topics first.Studies form topics using publication relations, including term co-occurrence, citation links, and hybrid approaches.Source: Establishing Research Topics · p. 3
- Topic modeling offers an alternative to clustering publications: each publication receives a probability distribution over generated topics.Source: Establishing Research Topics · p. 3
Bibliometric Indicators for Identifying Emerging Research Topics
Bibliometric indicators operationalize different attributes of emergence, but prior work varies in its definitions and scope. The paper argues for a more precise definition and a clearer connection between that concept and identification indicators.
Bibliometric Indicators for Identifying Emerging Research Topics
- Prior indicators track publication growth, core documents, citation positions, new terms or keywords, and changes in author numbers.Studies often combine indicators; one example combines keyword frequency, author counts, and reference interdisciplinarity.Source: Bibliometric Indicators for… · pp. 3–4
Further Discussion on Related Work
- Definitions of emergence lack consensus: proposed attributes range from growth and novelty to economic potential, science-basedness, and impact.Rotolo et al. (2015) criticize approaches that emphasize novelty and growth while neglecting other potentially important attributes, such as impact.Source: Further Discussion on Related Work · p. 4
- Many studies use small, field-specific databases or predefined journals and keywords, limiting their coverage of topics relevant worldwide.Source: Further Discussion on Related Work · p. 4
- Only a few studies use large-scale databases; ERACEP matches clusters across time slices, while Small et al. (2014) combine direct-citation and co-citation clustering.ERACEP identifies three emerging-topic patterns; the Small et al. approach uses novelty and growth.Source: Further Discussion on Related Work · p. 4
- Cluster matching may improve precision and coherence, but it also makes identification approaches more complex and less transparent.Source: Further Discussion on Related Work · p. 5
- The study adapts Rotolo et al. (2015), defines emerging research topics more precisely, and links the concept to indicators for identifying them.Source: Further Discussion on Related Work · p. 5
Methodology
The paper adapts Rotolo et al.’s emerging-technology definition for research topics, replacing societal and economic impact with scientific impact. Its proposed definition retains novelty, growth, and coherence, and treats uncertainty as less relevant because scientific impact should appear soon after emergence.
Definition of Emerging Research Topics
- Rotolo et al. (2015) define emerging technologies through novelty, growth, coherence, impact, and uncertainty, but technologies and research topics are distinct concepts.Source: Definition of Emerging Research Topics · p. 5
- For research topics, prominent impact refers to scientific rather than societal and economic impact.Source: Definition of Emerging Research Topics · p. 5
- The study considers uncertainty less relevant because it requires prominent scientific impact within a relatively short time after a topic emerges.It reasons that researchers may present work at conferences or workshops before publication and database indexing.Source: Definition of Emerging Research Topics · p. 6
- An emerging research topic is defined as radically novel and relatively fast growing, with a degree of coherence and considerable scientific impact.The resulting attributes are radical novelty, relatively fast growth, coherence, and scientific impact.Source: Definition of Emerging Research Topics · p. 6
Assigning Publications to Research Topics
The study builds research topics with direct-citation clustering and evaluates them using four emergence criteria. Its graphene example qualifies under one threshold set but not a stricter growth threshold.
Assigning Publications to Research Topics
- Topics are formed using direct-citation clustering; the analysis covers about nine million Web of Science articles and reviews from 2003–2012.The study uses the Waltman and Van Eck (2012, 2013) clustering method.Source: Assigning Publications to Research… · p. 6
Identifying Emerging Research Topics
- Growth is assessed using smoothed annual publication counts rather than raw counts to correct for random fluctuations.The smoothing averages counts over three consecutive years.Source: Identifying Emerging Research Topics · p. 7
- The criteria require rapid growth, early-stage novelty, scientific impact, and coherence, with each attribute tied to its own threshold.Growth, novelty, impact, and coherence are specified by r_i,t,Δt ≥ r_min, p̄_i,t ≤ p_max, c_i,t,Δt ≥ c_min, and h_i ≥ h_min, respectively.Source: Identifying Emerging Research Topics · p. 8
- Novelty, impact, and coherence must satisfy their criteria for the same year; a topic may have multiple emergent phases, but only its first is reported.The first period is recorded for convenience in reporting results.Source: Identifying Emerging Research Topics · p. 8
- With r_min=5, p_max=100, c_min=2,500, and h_min=1, graphene qualifies as emerging; with r_min=10, it does not.Whether a topic qualifies therefore depends on the parameter values.Source: Identifying Emerging Research Topics · p. 9
Clustering Results
The analysis clusters about 10,000 publications, then applies two parameter sets to identify emerging research topics. The sets identify 16 and 15 topics, with 12 shared.
Clustering Results
- Around 10,000 clusters were obtained, averaging 956 publications each; small clusters were retained because they had little influence on subsequent analysis.Source: Clustering Results · p. 9, Figure 2 · p. 9
Identified Emerging Research Topics
- For each cluster, growth, novelty, scientific impact, and coherence were measured; growth and citation impact use Δt windows, while coherence is within-cluster citations divided by publications.Source: Identified Emerging Research Topics · p. 10, Table 2 · p. 10
- Set 1 requires growth ≥2 and ≥1,500 citations over two years; set 2 requires growth ≥5 and ≥2,500 citations, with both requiring ≤100 first-year publications and coherence ≥1.Source: Identified Emerging Research Topics · p. 10, Table 3 · p. 10
- The two parameter sets identified 16 and 15 research topics, respectively, with 12 identified by both.Source: Table 3 · p. 10, Identified Emerging Research Topics · p. 10
- The strict parameter values yielded around 15 topics, though more may exist in the 10-year period; the precision–recall trade-off cannot be quantified without a definitive list.Source: Identified Emerging Research Topics · p. 11
Evaluation Tests
The evaluation faces no definitive validation standard, so the study emphasizes transparent criteria and uses topic demonstrations and comparisons to assess how its results cohere with prior analyses.
Evaluation Tests
- Validation by systematic comparison is constrained by the absence of gold standards, varying definitions and attributes, and different aggregation levels across studies.The authors note that aggregation choices obstruct comparisons with other methods.Source: Evaluation Tests · p. 11
- Expert validation is difficult because the analysis spans the science system and experts may disagree about emergence and the paper’s definition.Such differences could make expert accuracy criteria diverge from the study’s indicators.Source: Evaluation Tests · p. 11
- The authors therefore emphasize transparent, straightforward indicators aligned with their emergence definition and use prior-work comparisons for coherence insights, not definitive validation.The evaluation also includes an in-depth analysis of library and information science.Source: Evaluation Tests · p. 12
Demonstrating the Proposed Method
- Topic 1, covering graphene’s electronic properties, is detected under both parameter sets and meets the study’s novelty, growth, coherence, and impact attributes.Its growth rate is 5.24 from 2005 to 2007 and 34.05 over five years; the respective citation counts exceed 6,000 and 100,000, and cluster coherence is 21.Source: Demonstrating the Proposed Method · pp. 12–13, Table 4 · p. 12
- Topic 13 is detected with one parameter set but not the other because its five-year growth rate of 2.28 falls below r_min=5.The topic concerns EML4-ALK fusion gene studies in non-small cell lung cancer and shows a sharp publication increase from 2010 to 2012.Source: Demonstrating the Proposed Method · pp. 14–15
- Table 4 and Figure 3 report the emergence attributes and publication trends for the other identified topics, which the authors consider reasonable to regard as emerging.The four attributes are radical novelty, relatively fast growth, coherence, and scientific impact.Source: Demonstrating the Proposed Method · pp. 12, 15, Table 4 · p. 12, Figure 3 · p. 14
Comparison with the Previous Studies
The comparison offers context for how identified topics align with prior analyses, but is not intended as a systematic test of methods or accuracy.
Comparison with the Previous Studies
- The comparison is intended to show how the results cohere with previous analyses, not to systematically compare methodological effectiveness or accuracy.The authors caution that such comparisons may not provide sufficient evidence to validate their results.Source: Comparison with the Previous Studies · p. 15
- Small et al. (2014) identify eight emerging topics with “graphene” in their labels, while their micro-level aggregation may split the present study’s graphene topic into several clusters.The present study uses Topic 1 as an example of graphene research.Source: Comparison with the Previous Studies · p. 15
- Only a few prior works identify emerging topics globally; ScienceWatch’s related indicators have insufficiently clear definitions and methods to distinguish their terms.The comparison identifies Small et al. (2014) and ScienceWatch as the two main global references.Source: Comparison with the Previous Studies · p. 15
In-depth Analysis for the Field of LIS
The LIS analysis tests the proposed criteria on nine LIS-related clusters using two time intervals. Topic 90 emerges as the field’s most notable case, although it does not satisfy the study’s criteria under either parameter set.
In-depth Analysis for the Field of LIS
- The analysis selects nine clusters with more than 50% of publications in LIS and examines emergence attributes at Δt=2 and Δt=5.The field is used as a familiar example to examine whether the criteria produce meaningful results.Source: In-depth Analysis for the Field of LIS · p. 17
- At Δt=2, Topics 90, 7266, and 9354 only slightly exceed the mean maximum growth rate of 1.34; none meets the fast-growth parameter r_min=2.The analysis reports no LIS-related topic as relatively fast-growing at this interval.Source: In-depth Analysis for the Field of LIS · p. 17
- At Δt=2, Topic 90 has relatively strong scientific impact but is not considered radically novel because its starting-year publication count exceeds 200.The supplied Table 6 passage identifies the table as statistics on LIS-related research topics.Source: In-depth Analysis for the Field of LIS · p. 18, Table 6 · p. 18
- At Δt=5, Topic 90’s growth exceeds the 90th percentile (r=2) and its scientific impact is strong, but 327 publications in 2005 exceed the novelty threshold.The three topics identified at Δt=2 also retain growth rates above the mean at Δt=5.Source: In-depth Analysis for the Field of LIS · p. 18
- Topic 90 is judged the most emerging LIS topic over the 10-year period despite failing both parameter sets, indicating that values should reflect research purpose and field.It concerns citation use for research evaluation; the H-index’s influence is cited as context for its significance.Source: In-depth Analysis for the Field of LIS · p. 18
Discussion and Conclusions
The paper proposes a transparent, flexible bibliometric approach that links an explicit definition of emergence to indicators. Its application is bounded by data coverage, clustering coherence, and the need to tailor parameters and aggregation to the field.
Discussion and Conclusions
- The approach applies criteria to topics built from direct citation relations, identifies topics using two parameter sets, and evaluates them through demonstrations, prior studies, reports, and an LIS analysis.The method elaborates the definition of an emerging topic and connects emergence attributes to operational indicators.Source: Discussion and Conclusions · p. 19
- Parameter values and aggregation should suit the field and research purpose; social-science topics may call for finer aggregation and relaxed growth restrictions.Users must agree with the paper’s definition of emergence before applying the approach.Source: Discussion and Conclusions · p. 19
- The analysis uses WoS articles and reviews, so the method may be less useful where monographs or conference papers carry novel knowledge; database indexing fluctuations may affect accuracy.The paper notes conference-centered frontier research as a possible reason few computer-science topics were identified.Source: Discussion and Conclusions · p. 19
- The clustering method does not ensure sufficiently coherent clusters, so a coherence-evaluation criterion is necessary.The paper also identifies excluding small clusters or merging them with larger clusters as alternatives.Source: Discussion and Conclusions · p. 20
Appendix
The appendix includes LIS-related topic labels and lists of frequent terms, alongside a table identified as giving topic labels and their two most-cited publications.
Appendix
- The term lists include topics associated with resistive switching and photoelectrochemical water splitting.Source: Appendix · p. 22
- Other listed terms identify cognitive radio and transformation optics topics.Source: Appendix · p. 22
- Table A2 is identified as listing LIS-related topic labels and the two most-cited publications for each topic.Source: Table A2 · p. 23