Sharing the same bed with different dreams: Topic modeling the research-practice gap in public relations 2011-2020

ABSTRACT

Prior empirical efforts in uncovering the research-practice gap in public relations have often been restricted to perceptions and evaluations of people participating in the investigation. Moving beyond the linear perspective on knowledge transfer that dominates relevant discussions for decades, this study adopted topic modeling as an inductive analytical approach to examine a comprehensive set of texts representing the perspective of scholars and practitioners over a 10-year period from 2011 to 2020. A comparison of 35 topics discerned from academic journals (1,209 titles/abstracts) and professional texts (2,378 articles) revealed that a total of 18 topics were peculiar to each corpus, providing sound evidence of the substantial divide between scholars and practitioners. However, two communities shared common or comparable concerns over 17 topics, suggesting a significant convergence on crucial issues. Moreover, scholars and practitioners assigned varying weights to these topics in their publications, which indicated noteworthy differences in the primary areas of interest for both communities. In addition to deepening our understanding of the width and nuances of the research-practice gap in the field of public relations in a quantitative way, findings obtained from this study also signal the direction toward which scholars and practitioners should make progress to bridge the gap.

KEYWORDS:

• Knowledge community
• practitioner
• public relations
• research-practice gap
• scholar
• topic model

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the corresponding author, Xiao Wang, upon reasonable request.

Notes

1. Specifically, PRS is a quarterly magazine aiming at executive-level PR professionals with insightful feature-length commentary on the strategic importance of PR at the management level and views on changing PR concepts to engage and inform professionals and their organizations; while PRT is a monthly tabloid newspaper conveying essential tips shared by frontline PR practitioners on a wide range of topics and aims to empower communicators to advance their careers and showcase the positive influence of the profession.

2. Titles and abstracts included in the research corpus were retrieved from the official website of PRR (https://www.sciencedirect.com/journal/public-relations-review) and JPRR (https://www.tandfonline.com/loi/hprr20), respectively; articles for the S&T were collected from the publication’s online archive (https://www.prsa.org/publications-and-news/strategies-tactics). Both academic journals require authors to provide a concise and informative title serving as the abbreviated version of the major idea of their articles. Moreover, the abstract should be able to stand alone to briefly demonstrate the research purpose, primary findings, and principal conclusions. The retrieval of title and abstract information can ensure capturing the major idea of journal articles. On the other hand, the magazine-style pieces of S&T necessitate reading through the full text of each article to comprehend its main idea. Hence, the research corpus and the practice corpus scrapped for topic modeling are considered comparable to each other.

3. Prior studies suggest that Dirichlet hyperparameters α and β should be used to control the concentration/sparsity of topic distribution over documents (i.e., document-topic distribution) and word distribution over topics (i.e., topic-word distribution), respectively (Gerow et al. 2018). Lower values of hyper-parameters often result in a higher concentration of document-topic or topic-word distribution. These efforts are primarily made upon a single corpus to reach a balanced word distribution over topics. For the comparative purpose contained in the present study, however, controlling for β to manually set restrictions to the topic distribution is likely to produce a coerced result. Therefore, we fitted the LDA model with α only.

4. In practice, 1 is often used as the upper threshold for the fine tuning of α, particularly when there is prior knowledge that the distribution of topics over documents would be sparse and only a very few topics would be contained in a document (Naushan 2020). In the present study, however, it is likely that each document, from either research corpus or practice corpus, contains multiple topics. In addition, the data used for this study covers a 10-year period of observation comprising a variety of themes and viewpoints. It is thus more reasonable to use the common default of α (i.e., 50/K = 1.43) to obtain a set of more explicitly differentiated topics over documents.

Additional information

Funding

This work was supported by the GESIS (Leibniz Institute for the Social Sciences) under Grant [number GG-2019-017].

Notes on contributors

Xiao Wang

Xiao Wang received his PhD from The Chinese University of Hong Kong and is currently an Assistant Research Fellow in the School of Journalism and Communication at Nanjing University. His research interests lie in risk and crisis communication, marketing communication, and evolution of emergency collaborative networks.

Maggie Mengqing Zhang

Maggie Mengqing Zhang is a PhD student at the Institute of Communication Research, University of Illinois at Urbana-Champaign. Her research interests include computational marketing communication and political communication, with particular focuses on communication process and media effect.