Sage Journals: Discover world-class research

Abstract

Open science movement gains attention since it might enable a second scientific revolution that fundamentally changes research methods and standards across science. However, the discussion topics towards opens science both from the academia and the environments outside the scholarly communication process have not been formally identified. This paper contributes to that end by analyzing 145,716 open-science-related tweets and 3,200 research papers in Scopus from 2011 to 2022. The results show there is increasing interest about open science both on Twitter and from academia. There are similar foci for both the public on Twitter and the academia from Scopus, including cloud computing and COVID-19 pandemic. When the public on Twitter focus on open science events and citizen science, the scholarly research is more concerned about the detailed aspects and novel innovation in research. The findings might interest the policy-maker for offering evidence to facilitate open science policies and practices.

Keywords

open science debate temperature metrics bibiometric analysis

Introduction

Over the past years, science has witnessed a shift towards openness, transparency, and reproducibility, a movement known as “open science” (Bartling & Friesike, 2014). Open science is a umbrella concept and implies the opening of all phases of scientific research, as well as participatory process for determining the scientific and research agenda in relation to the public and their concerns (Nosek et al., 2015). The open science movement continues to gain momentum, attention, and discussion since it might enable a second scientific revolution that fundamentally changes research methods and standards across science, acknowledging the rapid technology changes primarily driven by the Internet (Choi, 2023; Homolak et al., 2020; Lefebvre & Spruit, 2023; Ramachandran et al., 2021).

Open science seeks new modes of relationship between the traditional creators of knowledge, such as scholarly researchers, and a wide range of users and those interested in it (Rodriguez-Pomeda et al., 2023). Hence, open science could be interested to not only those primary researchers, but also the general public (Voytek, 2017). Particularly, the challenge of the public health emergency, such as the COVID-19 pandemic gave evidences that open access to the critical scientific information and material is both crucial to the general public and the researchers (Besancon et al., 2021; Boby et al., 2023; Molldrem et al., 2021). As a potential public sphere (Dursun & Yildiz, 2022), social media platforms facilitate people to reflect and express opinions, so that could be connections between monotonic scholars and the general public (Y. Zhang et al., 2024). It is regarded that social media provide a platform to run “biggest research conference in the world” instead of offering costly affairs only available to the privileged relative few (Voytek, 2017). Zong et al. (2023) adopted the social media attention on articles to estimate the effects of open science badges and at last the effectiveness of the open science policy.

As a influential social media, Twitter is often adopted as an empirical source for scientific research (Brembs et al., 2023; Karami et al., 2020; L. Zhang et al., 2023). It is regarded that microblogging on Twitter extends public science communication by providing additional voices and directing attention (Buchi, 2017). Also, Twitter has been found to facilitate direct interaction between researchers and the public (Cheplygina et al., 2020). Hence, Twitter could be employed side-by-side with citations as “altmetrics” (Jia et al., 2020). Fang et al. (2022) analyzed scientific articles and academic tweets to focus on the user engagement behavior around scholarly tweets. Compared to other four open altmetrics data sources including Mendeley, News, Blogs and Policy, Twitter attention surrounding research output both starts and ends quickly (Taylor, 2023). Besides technologies, the impact of publications in the field of social sciences could also be evaluated adopting Twitter as open altmetrics (Sedighi, 2023).

In addition to act as new generation metrics for research quality and impact, Twitter is also adopted in public opinion mining on specific science-related topics. To understand patron engagement with the library, Stewart and Walker (2018) analyzed retweets and Twitter followers to identify that most tweets were related to “institutional boosterism.” One study examined Lanadian tweets on Marijuana legalization and terminology to demonstrate how this kind of research method may be used to inform library practice (Kung et al., 2024). DORA (Declaration on Research Assessment)-related tweets were collected to identified the viewpoints on social media (Orduna-Malea & Bautista-Puig, 2024). The open access movements were also examined on Twitter for its main features (Sadiq & Yadav, 2022; Sotudeh, 2023; Sotudeh et al., 2022). All the above research works demonstrate that Twitter is appropriate for both open altmetrics and opinion mining for science-related topics, but few of them formally identified the discussion topics and trends about open science on Twitter.

Besides opinion mining on Twitter, topic modeling is popular in mining the voice of the academic from research publications (Wang et al., 2023). In many cases, the academic view counts a lot in policy-making, especially in the development of some special disciplines (Luhmann et al., 2022; Walsh et al., 2022). Because of the importance of bibliometrics study in academic research, there are continuous publications of the literature focusing on the important facets. Bashar et al. (2023) found the influence of COVID-19 on consumer behavior as a guide for the researchers and decision-makers to strategies accordingly through a bibliometric review analysis. A bibliometrics study of the cancer research in the faculties of medicine and dentistry was reported in Al-Raeei et al. (2023) in order to investigate the scientific research outputs of the Damascus University. Hence, a comprehensive study is required to collect and analyze for both the public opinions on social media and the academic views studies to identify open science topics formally. Such comprehensive visions could act as public engagement for policy-maker that provide opportunities not to rethink their policies and practices, but to gain trust for a predetermined approach (Alexopoulos et al., 2014; Thorpe & Gregory, 2010).

For this aim, in this paper we raises the following research questions:

RQ1: What is the public opinion concerned about the umbrella concept “open science” on Twitter ?

RQ2: What kind of viewpoints does the academia hold towards “open science”?

RQ3: What is the consistency or difference between the public opinions and the academic view about “open science”?

These three research questions aim to consider open science both from public opinions on Twitter and the academic view in research publications. The first questions reflects the online public’s beliefs and focus on open science topics on Twitter. The second question describes the scholarly viewpoints the researchers hold for open science in their studies. The third question shows the consistency or differences between public opinions and academic views towards open science.

Data and Methodology

Research Design

To understand open science in a comprehensive way, the paper provides insights into both the tweets data and the Scopus data. A research framework shown in Figure 1 is designed to characterize the viewpoints from both the public and academic towards open science. Since both the public opinions and academic view towards open science are mainly objective statements, there were very few descriptions about emotion. And most of the tweets and research publications expressed support for open science.

Figure 1.

The research framework of mining topics on open science.

As shown in Figure 1, the main procedures in this research included three steps: data collection and cleaning, data processing, and data analysis. In the first step, the raw data from Twitter and Scopus database were collected and cleaned. The tweets data was collected through the hashtag “open science,” and was cleaned by getting rid of noise, stopwords and POS (part-of-speech). While the Scopus data was sourced from the Scopus database with the keyword “open science.” Duplication deletion was applied on the Scopus data to get rid of the repetition.

The second step was data processing. The cleaned data from both data sets were processed with topic modelling, respectively. The temperature metrics was adopted on tweet data for annual hot spot. And the VOSviewer application was employed to visualize the keyword co-occurrence of the Scopus data. In this paper, topic modelling adopted latent topic analysis to model and aggregate when figuring out the major topics in tweets or research publications, respectively.

In the third step, we employed different data analysis methods on tweet data and Scopus data, and finally made the comparison analysis between the two datasets. This study observed the major topics and diachronic changes on open science. Additionally, keyword co-occurrence could help to clarify and understand the focus on open science from researchers. The comparison between the viewpoints both from the public and academia could provide evidences supporting to policy-maker and participator in future, and eventually help facilitate open science practice.

Tweets Dataset

To test our proposed methods, we used a tweets dataset collected from Twitter. This dataset used “open science” as hashtag to crawl open-science-related conversations on Twitter. After data cleaning, we have collected the tweets data from January 2009 to December 2022. It could be observed from Figure 2 that open-science-tweets in 2011 were about 2000 posts. In 2009 and 2010, the tweets collections were only 103 and 490. These “open science” tweets rose to 6180 in 2012 and keep increasing till 2017. There is a slight decline after 2018 and the tweets collection end in 12549 tweets in 2022. Since the tweets collections in 2009 and 2010 are not large enough for data analysis, we only used the data from January 2011 to December 2022. At last, we totally got 145,716 posts on “open science.” These posts were made of notices, advertisements, opinions, etc. Since there were few retweets and likes, we ignored them and only examined the posts on Twitter.

Figure 2.

The change of the total tweet number on open science over year.

Scopus Dataset

We collected the published publications about “open science” from the Scopus database. We focused on the period from 2011 to 2022, which witnessed the bloom of the open science practice. The keyword “open science” was searched in the TIT-ABS-KEY field of Scopus search engine. Only articles and conference papers written in the English language were included. Totally 3,477 papers, including 2,649 articles and 828 conference papers, were returned in .CSV format. After deleting the duplication through checking the DOIs, we got 3,200 papers.

From Figure 3, we can see that the number of published papers increased from 41 in 2011 to 752 in 2022. The number of the published papers increased in a nob linear functionality and the increasing of the published papers is about 19 times. The increasing reflects the blooming of “open science” in academic. In Figure 4 and based on the Scopus data, we illustrated the published documents that are abstracted and indexed in Scopus database. It could be observed from Figure 4 that the published papers in computer science represent about 18.2% of the total published papers towards “open science,” and are the most. While the published papers in social science represent about 14.9%, are the second. Also, the third are the published papers in medicine, which count for 10.4% of the total published papers towards “open science.” This proportion reflects the various interests towards “open science” in different subject areas.

Figure 3.

The “open science” papers abstracted and index in Scopus database from 2011 to 2022.

Figure 4.

The proportions in different subject areas.

Temperature Metrics

In order to identify the focus of open science on Twitter each year, we need to find the representative keyword. Previous works (Troussas et al., 2019; S. Zhou et al., 2020) preferred high-frequency words as hot words to represent the hot spots. However, in this paper the hot words were almost the same for each year so that they couldn’t describe the changes of the viewpoints on open science over time. The reason for it is that experimental data were short text on various topics. Hence, there were a large number of low-frequency words and the high-frequency words tend to be fixed on some special words all the time.

To tackle the sparsity problems of the tweets data, we introduced temperature parameter to get more diversified results. The temperature parameter is a hyper-parameter used in language models such as GPT (Brown et al., 2020) to control the randomness of the generated text. It controls how much the model should take into account low-probability words when generating the next token in the sequence. The parameters control the degree of randomness or creativity in the generated output. Randomness could be regarded as the diversity of responses to multiple inquires. A high randomness might result in more creative responses, that is, a higher likelihood of answers without factual basis. On the contrary, low randomness could be described as multiple inquires being more likely to encounter repeated answers, while these answers are closer to fact, that is, closer to training data. The original metric was given in equation (1):

softmax (y_{i}) = \frac{e^{\frac{y_{i}}{T}}}{\sum_{j = 1}^{n} e^{\frac{y_{i}}{T}}}

(1)

Where $Y = (y_{1}, y_{2}, . . ., y_{n})$ is an input vector and values $y_{i} (\bar{1, n})$ are in range from $- \infty$ to $+ \infty$ , . The temperature parameter T can take on any numerical value. Zero temperature is equivalent to argmax Likelihood, while infinite temperature is equivalent to uniform sampling. That also means a lower value of the temperature parameter will lead to a more predictable and deterministic output, while a higher value will produce a more random and surprising output. The Temperature parameter is typically used to adjust the output of the softmax function, to increase or decrease the model’s confidence in different categories. Specifically, the softmax function converts the model’s predictions for each category into a probability distribution. The temperature parameter can be seen as a scaling factor that can increase or decrease the confidence of each category in the softmax function output.

In this paper we modified the original formulation to form our temperature metrics applied to the tweets and Scopus data. At first, we defined the probability of word w_i appearing in the dataset in the jth year as p_ij:

p_{ij} = \frac{\frac{t f_{ij}}{N_{j}}}{\sum_{j = 2011}^{2022} \frac{t f_{ij}}{N_{j}}}

(2)

Where tf_ij is the term frequency of w_i in the dataset in the jth year, and N_j is the number of all the documents in the jth year. Here, j is from 2011 to 2022. Then, the temperature metrics of w_i in the jth year is defined as:

temp (p_{ij}) = \frac{e^{\frac{p_{ij}}{T}}}{(\sum_{j = 2011}^{2022} e^{\frac{p_{ij}}{T}}) - e^{\frac{p_{ij}}{T}}}

(3)

Here, we adopted $(\sum_{j = 2011}^{2022} e^{\frac{p_{ij}}{T}}) - e^{\frac{p_{ij}}{T}}$ instead of $\sum_{j = 2011}^{2022} e^{\frac{p_{ij}}{T}}$ to highlight the difference between the jth year and the other years. For example, suppose p_ij=1 in jth year and p_ij = 0 in other 11 years. When adopted equation (1), we can get $softmax (p_{ij}) = \frac{e^{\frac{1}{T}}}{11 e^{0} + e^{\frac{i}{T}}}$ . If $T \to 0$ , $lim_{T \to 0} \frac{e^{\frac{1}{T}}}{11 e^{0} + e^{\frac{i}{T}}} \approx 1$ . But if we use equation (3), we can get $temp (p_{ij}) = \frac{e^{\frac{1}{T}}}{11 e^{0}}$ . If $T \to 0$ , $lim_{T \to 0} \frac{e^{\frac{1}{T}}}{e^{0}} \to + \infty$ . Hence, adopting equation (3) could amplify the difference between the target year and the other years.

Finally, the weight of the word w_i appearing in the jth year was:

S_{ij} = t f_{ij} \times temp (p_{ij})

(4)

In this paper, we adopted S_ij instead of term frequency to select the representative keywords for each year. The words with high S_ij will be chosen as the representative keywords. The temperature metrics allowed us to control the deterministic factors. When T is almost 0, the value of temp(p_ij) becomes positively infinite. On the other hand, the infinite temperature will bring 1/11 to temp(p_ij) in each year so that the representative keywords with high S_ij will be almost the hot words with high term frequency.

Topic Modelling and Aggregation

As an unsupervised machine-learning approach for discovering latent hidden semantic topics in large collections of documents, topic modelling could identify clusters of documents by a representative set of words. To understand the major topics being discussed in the data collections, this paper adopted Latent Dirichlet Allocation (LDA) proposed by Blei et al. (2003) as topic modelling techniques. The most highly weighted words in each cluster provide insight into the content of each topic. LDA requires users to input the number of expected topics. To determine the optimal number of topics, coherence score was adopted in this paper.

In the context of documents analysis, LDA has been criticized for producing non-replicable results by Steyvers and Griffiths (2007). Hence, aligning topics from multiple models was mentioned to ensure the reliability of LDA results. Many works have introducing approaches to increase the reliability of LDA (Blair et al., 2020). We took up this idea by adopting topic aggregation in our methods. When setting k topics in LDA model, we run n LDA models based on the previously determined hyperparameters and differing random states. This would provide n times k, that is, n × k topics. These topics are then clustered using k-means clustering, based on the cosine distance between their term probability distributions. In order to get rid of noise, we chose to perform topic modelling exclusively on nouns, proper nouns, and noun phrases. Since all function words are generally not relevant for identifying topics, the elimination provides noise reduction. As discussed in Martin and Johnson (2015), a nouns-only approach leads to better interpretable topics.

Assigning Themes in Public Opinions

There are several topics after LDA modeling on tweets data. Through discussions, the authors then grouped the topics of public opinions into broader themes. Our procedures are consistent with similar studies that have examined social media data using text mining and topic modeling (Chandrasekaran et al., 2020; Q. Zhou & Jing, 2020). In order to classify the open science topics, we adopted the classifications described in UNESCO (United Nations Educational Scientific and Cultural Organization) Recommendation on Open Science (UNESCO, 2021). In this Recommendation, UNESCO provide four open science themes: open scientific knowledge, open science infrastructures, open engagement of societal actors, and open dialogue with other knowledge systems. Besides the classifications described in UNESCO Recommendation on Open Science, there are still some ambiguous topics which are not clearly defined in UNESCO Recommendation. We classified these non-UNESCO topics into “others.” Since there were no topics about the theme “open dialogue with other knowledge systems,” we haven’t taken this theme into account in this paper.

Table 1 shows the tweet samples in the four themes in this paper. From these samples it could be observed that the tweet data covered various types of posts, including public opinions about open access, advertising of the projects or tools, testing of the open infrastructure, or open science in other disciplines. The intensive discussion on social media reveal the focus of the public on open science so that it could provide abundant material for policy maker to adopt appropriate open science policies.

Table 1.

Samples on the Four Open Science Themes.

Themes	Tweet sample
Open scientific knowledge	Periodic reminder to my fellow academics: Now if you publish with Elsevier, researchers in Sweden, Germany, or the UC network cannot access your papers. In fact, even if you share the “free access” link for authors to distribute, access from a Swedish institute is blocked.
Open science infrastructures	Testing. Try our cloud accounting blogs in one of European open science cloud.
Open engagement of societal actors	Crowdfunding is not over yet.
Others	OpenScience Promoting scientific integrity through open science in health psychology.

When grouping the topics of LDA into broader themes, we adopted the weights provided by LDA and a self-building theme vocabulary. The self-building theme vocabulary consists of words from the UNESCO Recommendation, as well as the tweets data set we collected. Table 2 shows the sample of the self-building theme vocabulary.

Table 2.

Self-Building Theme Vocabulary.

Themes/sub-themes		Words (three for brief)
Open scientific knowledge	Scientific publication	Paywall; science journalism; preprint
	Open data	Panton; data management; data sharing
	Open tool	Open source; open hardware; github
	Open education	Mooc (massive open online course); oer (open education resource); fellowship
Open science infrastructures		Cloud; h2020 project; Openaire
Open engagement of societal actors		Public engagement; community; citizenscience
Others	Open science events	Oaweek; okfest; opencon
	Open science definitions	UNESCO recommendation; ciencia; wissenschaft
	Open evaluation	Metrics; citation; impact
	Blockchain	Blockchain; bitcoin; protocol
	Discipline	Neuro science; machinelearning; climate
	COVID-19	Pandemic; vaccine; immunology

In the processing of the topic modelling, LDA will give the value of probability p (w|t) for the word w in topic t. Suppose T_m is the word set of the mth theme/sub-theme of the self-building vocabulary. Here, $m \in [$ 1, l], l is the total number of the theme/sub-theme in the self-building vocabulary. For a word w_i in the topic t_k, if $w_{i} \in T_{m}$ , we can get the weight of T_m to t_k as:

U_{m, k} = \sum_{i = 1}^{| t_{k} |} p (w_{i} | t_{k})

(5)

When grouping the topic to the themes, we assigned the T_m to t_k if U_m,k is the largest to t_k.

After assigning the themes/sub-themes, we selected the representative keywords for the themes/sub-themes. For the word w_i, $w_{i} \in T_{m}$ , we could get the weight V_i,m:

V_{i, m} = \sum_{j = 2011}^{2022} p_{j} (w_{i} | t_{k})

(6)

Here, p_j(w_i|t_k) is the p(w_i|t_k) in jth year. Then, we could get the values of V_i,m in a descending order and select the top 5 keywords as the representative keywords for the mth theme.

Results Analysis on Tweets Data

Annual Hot Spots Analysis

In this paper, we obtained the representative keywords when T = 0.3. This value of temperature could ensure the specificity of these representative keywords, as well as refuse trivial nonsense. From equation (3), it could be observed that the temperature metrics will highlights the difference among various years. Hence, it could be inferred that the annual hot spots selected through temperature metrics could reflect the specific focus of the public opinions for each year. From Table 3, it could be observed that when T = 0.3, we could get appropriate keywords. When T = 1.0, the keywords were too general to describe the specificity of each year. While when T = 0.1, the keywords were very particular that it might lack a macro view towards open science. The top 10 representative keywords for each year were shown in Table 3. To provide a overview of the annual hot spot on open science, we observed the corresponding tweets of the representative keywords and found the important events and entities. Also, a brief is given for the character of the public opinion towards open science each year.

Table 3.

Top 10 Keywords for Annual Hot Spot.

Year	Temperature	Top 10 keywords for brief
2011	T = 1.0	Science; opendata; openaccess; oss2011; research; mendeley; einstein; scientist; nielsen; share
	T = 0.3	oss2011; einstein; mendeley; okcon2011; science; genotype; nielsen; opensnp; opendata; scio11
	T = 0.1	oss2011; okcon2011; genotype; scio11; opensnp; tedxwaterloo; solo11; idcc11; einstein; superhero
2012	T = 1.0	Openaccess; science; opendata; research; okfest; access; frpaa; scientific; publishing; scientist
	T = 0.3	okfest; frpaa; oss12; openaccess; science; finchreport; sagecon; opendata; oamonday; pantonscience
	T = 0.1	Okfest; frpaa; oss12; sosea; finchreport; pantonscience; oaweek2012; macedonia; tornado; spektral
2013	T = 1.0	Science; openaccess; opendata; research; access; scientific; scientist; course; publishing; whchamps
	T = 0.3	Whchamps; schoolofopen; pdftribute; hyphdus; science; openoxford; openaccess; hyphd; opendata; rdalaunch
	T = 0.1	Whchamps; schoolofopen; grazierita; pdftribute; hyphdus; artículogratis; openoxford; hyphd; btpdf2; rdalaunch
2014	T = 1.0	Science; openaccess; opendata; research; opensource; scientific; sharing; access; project; journal
	T = 0.3	okfest14; science; opencon2014; esof2014; openaccess; caféos; opendata; esa2014; brainbrowser; f1000talks
	T = 0.1	okfest14; opencon2014; esof2014; osrio; caféos; esa2014; mswnews; oaweek2014; brainbrowser; f1000talks
2015	T = 1.0	Science; openaccess; research; opendata; opensource; great; access; journal; scientist; project
	T = 0.3	Eraofinnovation; science; ict2015; liber2015; openaccess; esa100; cambiare; opendata; research; opentherapeutics
	T = 0.1	Eraofinnovation; ict2015; liber2015; esa100; cambiare; oscarbadoino; saa2015; pcori2015; cspc2015; researchkit
2016	T = 1.0	Science; openaccess; opendata; research; access; researcher; great; eu2016nl; scientific; citizenscience
	T = 0.3	eu2016nl; science; openaccess; opendata; research; esof16; oaweek2016; cancermoonshot; digitiseeu; zikavirus
	T = 0.1	eu2016nl; esof16; oaweek2016; volta; visto; l’alieno; digitiseeu; force2016; zikavirus; sgctep
2017	T = 1.0	Science; research; openaccess; opendata; datascience; great; bigdata; scicomm; access; researcher
	T = 0.3	osfair2017; science; farming; smartfarming; openaccess; research; science4all; opendata; osc17; agriculture
	T = 0.1	osfair2017; defstar5; mpgvip; makeyourownlane; agrotes; smartfarming; osfair17; openscienceworks; oaweek17; force2017
2018	T = 1.0	Orvium; science; research; blockchain; openaccess; peerreview; publishing; crypto; opendata; author
	T = 0.3	Orvium; blockchain; ethereum; tokensale; bitcoin; cryptocurrency; science; peerreview; telegram; research
	T = 0.1	Orvium; crypto; tokensale; ether; ethereum; telegram; bitcoin; cryptocurrency; publishingcation; esof2018
2019	T = 1.0	Science; research; openaccess; opendata; project; great; researcher; opensource; access; scicomm
	T = 0.3	Science; research; filesfmlibrary; openaccess; osfair2019; oaweek19; osmooc; opendata; osc2019; project
	T = 0.1	Filesfmlibrary; osfair2019; oaweek19; osc2019; ohbm2019; oai11; oaweek2019; porto; oaspa2019; liber2019
2020	T = 1.0	Science; research; openaccess; covid19; opendata; coronavirus; access; covid-19; researcher; community
	T = 0.3	Coronavirus; covid19; science; research; covid-19; openaccess; pandemic; opendata; oaweek2020; researcher
	T = 0.1	Coronavirus; covid-19; oaweek2020; covid2019; careerassessment; rewardsystem; beyondcovidchallenge; osbiz2020; operas2020; 2019ncov
2021	T = 1.0	Science; research; openaccess; opendata; project; researcher; scicomm; community; access; journal
	T = 0.3	Science; research; inpst; osfair2021; phdvoice; openaccess; dhpsp; futureofopenscience; osc2021; scicomm
	T = 0.1	osfair2021; saclation; osc2021; orionfinalconf; reactjs; hospitalstalktolovedones; oaweek2021; futureofopenscience; eoscsymposium2021; osw2021
2022	T = 1.0	Science; research; openaccess; access; community; project; researcher; opendata; journal; bioinformatics
	T = 0.3	Science; research; rrids; bioinformatics; openaccess; useful; osec2022; womenwhocode; desci; community
	T = 0.1	Rrids; useful; osf2022de; methodsmatter; osf2022nl; osec2022; roadtoopenscience; oaweek2022; bibliohj23; fdo2022

2011 The Starting Year of Open Science: Some First-Time Discussion or Contest to Initiate Open Science

oss2011, okcon2011, scio11: conference and events. oss: Open Science Symposium. Okcon: Open Knowledge Conference. Scio: ScienceOnline meeting.

Einstein, nielsen: In the Wall Street Journal on October 31, 2011, Michael Nielsen reminded scientists that Sharing information is fundamental for promoting the progress of science and technology in the article “The New Einsteins Will Be Scientists Who Share: From cancer to cosmology, researchers could race ahead by working together-online and in the open.”

Mendeley, genotype, opensnp: The corresponding tweet is “Share your genotype: openSNP wins Mendeley/PLoS API Binary Battle.” On November 30, 2011 Mendeley along with PLoS held the first Binary Battle Apps for science contest to carry open science further. OpenSNP is a non-profit, open-source project that is about sharing genetic and phenotypic information.

2012 Open Science Policy Establishment: Some Rules and Act Established as Basic Blocks of Open Science

Okfest, oss12: conference and events. Okfest is the community-curated Open Knowledge event that brings together the global community of working groups, local groups and all manner of people interested in all aspects of openness.

sagecon, oamonday: The users’ needs for policies. In 2012, the Oregon Sage-grouse Conservation Partnership (SageCon) was convened at the request of the Governor’s office to formulate an “all lands, all threats” approach to sage-grouse conservation. On May 21st, 2012, the folks behind the open access advocacy site access2research.org unveiled the site and kicked off the push to petition the White House to allow public access to the results of taxpayer-funded research (OAmonday).

Frpaa, finchreport, pantonscience: Policies on open science. The keyword “frpaa” refers to “Federal Research Public Access Act,” a USA Congress act about open access. The keyword “finchreport” is the Finch Report, which contains the recommendations of the Working Group on Expanding Access to Published Research Findings. The “pantonscience” refers to Panton Principles, Principles for open data in science.

2013 The Boom of Communities and Events on Open Science

schoolofopen, hyphdus, hyphd: The communities on open science. “schoolofopen” is a global community of volunteers providing free online courses, face-to-face workshops, and innovative training programs on the meaning, application, and impact of “openness” in the digital age. “hyphdus&hyphd” refers to the open science community HackYourPhD, born in January 2013. HackYourPhD gathers PhD students and researchers to explore events and initiatives regarding new approaches of research.

Openoxford, Whchamps, rdalaunch, pdftribute: Events on open science. “openoxford” refers to open Oxford. “whchamps” is “White House champion of change,” which is for connected learning and all about library design. “rdalaunch” refers to the meetings of the Research Data Alliance (RDA). “pdftribute” was started in response to the tragic death of Aaron Swartz, who had contributed for open access.

2014 Tools and Services on Open Science

okfest14, opencon2014; esof2014, esa2014: conference and events. OpenCon is an annual regional event which celebrates the benefits of open education to student learning. Esof: EuroScience Open Forum. esa2014: 2014 ESA (Ecological Society of America) Annual Meeting.

Caféos: Open Science Cafes consists of round-table conversations that are ignited by statements on a set of cards.

Brainbrowser: BrainBrowser is an open source JavaScript library exposing a set of web-based 3D visualization tools primarily targeting neuroimaging.

f1000talks: F1000 provides open research publishing services to funder, institutions, and societies worldwide.

2015 the Beginning of the EU Cloud and Open Science Practice

ict2015, liber2015: conference and events. ICT2015 (Innovate, Connect, Transform) is the largest event organized by the EU Commission in 2015 to foster research and innovation at European level, to show the best results achieved through financed projects, to inform about upcoming call, and to facilitate the emergence of new networks. LIBER is the voice of European research libraries.

Eraofinnovation: Speeches and presentations from Opening up to an ERA of innovation.

esa100: openscience meeting. In December 2015, the Ecological Society of America celebrated the 100th birthday. The conversation continues on Twitter under the centennial hashtag #ESA100.

Cambiare: The corresponding tweet is “OpenScience, una sola rete cloud UE per cambiare la ricerca.” The tweet means a single EU cloud network to change research.

Opentherapeutics: Open therapeutics curates and develops open medical, biopharma, and synthetic biology-based biotechnologies.

2016 Practice on Open Science and Zika Virus

eu2016nl, esof16, oaweek2016: conference and events. eu2016nl: the EU Presidency Conference on Open Science 2016 at Netherlands. Oaweek: open access week.

Cancermoonshot: “cancermoonshot” refers to the event happened in 2016 that the USA president Obama leaded a task to help the end cancer.

Digitiseeu: EU Commission presented strategy on Digital Single Market and Open Science. Zikavirus: In 2016, WHO declared that the association of Zika infection with clusters of microcephaly and other neurological disorders constituted a Public Health Emergency of International Concern.

2017 Open Science Practice on Farming and Citizen Science

osfair2017, osc17: conference and events. Osfair: Open Science Fair, the inaugural international conference on Open Science. Osc: open science conference.

Farming, smartfarming, agriculture: open science practice.

science4all: “science4all” refers to Science4All.org, a website of quality popular science for educational purposes without any cost.

2018 Blockchain and Open Science

Orvium, ethereum: “orvium” and “ethereum” were all platform for open publish or open source based on blockchain technology.

Blockchain: technologies.

Tokensale, bitcoin, cryptocurrency: virtual currency based on blockchain.

Telegram: telegram group on open science and blockchain.

2019 Open Science Based on Blockchain and Online Course

osfair2019, oaweek19, osc2019: conference and events.

Filesfmlibrary: The keyword refers to files.fm, a website to help people upload, find and access useful content stored on user community supported storage platform. Files.fm data storage platform has received European Union support for practical implementation of blockchain solution.

Osmooc: open science massive open online courses (mooc).

2020 COVID-19

oaweek2020: conference and events.

Coronavirus, covid19, covid-19, pandemic: COVID-19.

2021 More Open Science Practice, Especially on Health

osfair2021, osc2021: conference and events.

Inpst: The International Natural Product Sciences Taskforce (INPST).

Phdvoice: an organization that’s committed to helping PhD students.

Dhpsp: The Digital Health and Patient Safety Platform (DHPSP).

Futureofopenscience: the discussion of the future of open science.

Scicomm: Sci-comm is the practice of informing, educating, and raising awareness of science-related topics. This often refers to communicating to non-specialists, as opposed to expert-to-expert communication associated with scientific publishing.

2022 More Diverse Open Science Practice, Including Different Infrastructure, Discipline, or Gender

osec2022: conference and events. Osec: Open Science European Conference.

Rrids: The keyword refers to “Research Resource Identifiers,” the resource identification initiative, developed and recommends to help improve reproducibility.

Bioinformatics: Bioinformatics, as related to genetics and genomics, is a scientific discipline that involves using computer technology to collect, store, analyze and disseminate biological data and information.

Womenwhocode: Women Who Code is on a mission to empower diverse women to excel in technology careers.

Desci: it refers to “Decentralized science,” a movement that aims to build public infrastructure for funding, creating, reviewing, crediting, storing, and disseminating scientific knowledge fairly and equitably using the Web3 stack.

Themes and Prevailing Trends

Through topic modeling, we identified various topics on open science in different years, respectively. The numbers of topics in different years are shown in Table 4. Figure 5 shows the proportions of each themes on total tweets data set. From Figure 5, it is obviously that “open scientific knowledge” is the hottest theme (49%), which attracted most attention from social media. The theme “others” also accounts for a large proportion (34%) in total data set because it covered various topics outside of the UNESCO Recommendation. The theme “open engagement of societal actors” (11%) and “open science infrastructures” (6%) were less intensively discussed compared to the former two themes.

Table 4.

The Number and Proportions of the Themes in Each Year.

Year	Total topics numbers	Open scientific knowledge (%)	Open science infrastructures (%)	Open engagement of societal actors (%)	Others (%)
2011	11	70.51	0	0	29.49
2012	15	63.01	0	6.15	30.84
2013	15	42.08	0	20.68	37.24
2014	16	70.69	0	11.91	17.4
2015	16	40.19	13.44	23.73	22.64
2016	15	58.24	6.36	16.93	18.47
2017	18	45.3	11.07	6.25	37.38
2018	15	48.42	5.93	8.94	36.71
2019	15	57.57	2.59	3.56	36.28
2020	16	31.21	6.33	3.04	59.42
2021	15	38.51	7.15	18.79	35.55
2022	15	53.83	3.99	2.9	39.28

Figure 5.

The proportion of the four themes on open science.

Table 4 shows the topic numbers and the proportions of the themes on open science each year, and Figure 6 describes the themes trends over time. From Table 4 we can observe that the topic numbers kept to increase until 2017, when the peak is 18. After that, there is a slight decline in the total topic numbers and it ended in 15 in 2022. The trend of the tweets topic numbers is consistent with the trend of the total tweets numbers shown in Figure 2. It is hence evident that the optimal numbers of topic modeling are reasonable.

Figure 6.

The change of the four open science themes over time.

From Table 4, it is observed that the theme “open engagement of societal actors” started at 2012 (6.15%) and boomed in 2013 (20.68%). Also, the theme “open science infrastructures” began at 2015, which is consistent with the annual hot spot analysis-there appeared the first tweet about the EU cloud in 2015. In addition, as shown in Table 4, it is worth noting that the development of the open science could not be separated from the boom of the innovative technologies. For example, the open science for virtual infrastructures in 2015 was based on the cloud computing starting in 2014 (Dordevic et al., 2014). Also, before the start of “open engagement of societal actors,” the keyword “open business” was mentioned in tweets, which is the predecessor of “crowdfunding” and “crowdsourcing.” These facts provide evidence that the development of open science is closely linked to the advance of technology.

From Table 4 and Figure 6, we could find that the theme “others” exceeded all the other themes on open science in 2020. The reason is that in 2020, COVID-19 broke out, leading to the explosion of the related topics. These tweets were mainly about influence the COVID-19 exerted on open science practice, or the benefits of taking open science events in the pandemic. The interactions between “COVID-19” and “open science” indicate that open science practice was closely bound up with the current events and users’ needs.

Theme 1: Open Scientific Knowledge

The theme “open scientific knowledge” refers to open access to scientific resources (UNESCO, 2021). According to the Recommendation, there are five sub-themes contained in this theme: scientific publications, open research data, open educational resources, opensource software, and source code, as well as open hardware. Since opensource software and source code, as well as open hardware are all open tools, in this paper, we combined the two sub-themes into the open tool sub-theme.

Figure 7 shows the proportions of the different sub-themes in the theme “open scientific knowledge.” It is obviously that “scientific publication” possesses the largest share (50%) in this theme. The sub-theme “open data” accounts for 25% in the theme, much lesser than “scientific publication.” The share of the “open tool” (13%) and “open education” (12%) is similar. It shows that “scientific publication,” also mentioned as “open access,” was the most concerning sub-theme in the theme “open scientific knowledge.”

Figure 7.

The proportion of the topics in the theme “open scientific knowledge.”

The five representative keywords in Table 5 describe the main contents of various sub-themes. It could be observed that these sub-themes contained different details on open science. “Scientific publications” is about “openaccess,” and users were more concerned about “article,”“preprint,”“journal,”“publish,” etc. “open data” include the science, rules, and visualization of data, such as “datascience,”“dataviz,” etc. This sub-theme also contains data itself, such as “bigdata” and “clinicaltrials” (clinical trials data). The sub-theme “open tool” was composed by “openhardware” and “opensource,” while “github” is a opensource community. The keywords for “open educational resources” were mainly “openscholarship,”“training,”“school,”“course,” etc.

Table 5.

Keywords Contained in Four Sub-Themes Belonging to Theme 1.

Sub-themes	Keywords (top 5 for brief)
Scientific publications	publish; preprint; journal; article; openaccess
Open research data	datascience; bigdata; clinicaltrials; dataviz; opendata
Open tool	openhardware; github; opensource; platform; project
Open educational resources	school; openscholarship; course; training; education

Theme 2: Open Science Infrastructures

The theme “open science infrastructures” refers to the sharing research infrastructures that are needed to support open science and serve the needs of different communities. It mainly includes virtual and physical infrastructures (UNESCO, 2021).

As shown in Table 6, these keywords could be classified into two classes: the virtual infrastructure, including “cloud” and “openaire”; and the supporting project, such as “h2020,” and “horizoneu.” Here, the keywords “openaire” is Open Access Infrastructure Research for Europe, an active network in 35 countries; “cloud” refers to the cloud computing platform. The keywords “h2020,”“horizon” and “horizoneu” all refer to the horizon project in Europe, the EU’s key funding program for research and innovation.

Table 6.

Keywords in Theme 2 and Theme 3.

Themes	Keywords (top 5 for brief)
Open science infrastructures	openaire; European; h2020; cloud; horizoneu
Open engagement of societal actors	citizenscience; crowdsourcing; crowdfunding; collaboration; project

Theme 3: Open Engagement of Societal Actors

The theme “open engagement of societal actors” is the extended collaboration between scientists and societal actors beyond the scientific community. They are mainly crowdfunding, crowdsourcing, scientific volunteering, and citizen and participatory science (UNESCO, 2021). The five representative keywords are shown in Table 6. In this theme, there were many projects and events of citizen science mentioned in the section “annual hot spot analysis,” such as “Cancermoonshot,”“sagecon,”“Openoxford,” etc.

Theme 4: Others

Besides the UNESCO classes, there are also some sub-themes which are not contained in the UNESCO Recommendation, as shown in Table 8. These sub-themes are open science events, open science definitions, open evaluation, blockchain, discipline, and COVID-19.

As shown in the Figure 8, “open science events” was the largest sub-theme (42%) in the theme “others.” The reason for that is because many organizations or authorities adopted Twitter as advertising platforms for their events. “open evaluation” accounts for 22% in the theme, shown that how to evaluate the performance of the open science projects or tools as well as the impact of the open access papers was most concerned by the public. The sub-theme “open science definition” (14%) describe the various definitions given to the umbrella concept “open science.” The sub-theme “blockchain” (10%) was mainly about the interaction between the blockchain and open science. The sub-themes “discipline” and “COVID-19” have the same share (6%), much smaller than the other topics.

Figure 8.

The proportion of the topics in the theme “others.”

From Table 7, it is obviously that open science practice exert important influence on the researchers’ life. In the sub-theme “open science events,” the keywords “conference,”“workshop,”“discussion,”“session,” etc., reflected that people usually communicate “open science” in these way.

Table 7.

Keywords for Sub-Themes in Theme 4.

Sub-themes	Starting year	Keywords (top 5 for brief)
Open science events	2011	conference; workshop; discussion; event; session
Open science definitions	2011	ouverte; abierta; ciencia; wissenschaft; research
Open evaluation	2012	altmetrics; peer review; evaluation; citation; factor
Blockchain	2017	orvium; crypto; bitcoin; blockchain; linkedresearch
Discipline	2017	space; digital health; bioinformatics; climate; open sociology
COVID-19	2020	pandemic; COVID-19; immunology; patient; raredisease

The keywords contained in the sub-theme “open science definitions” are all about the definition and classification of open science. Here, “ciencia” and “wissenschaft” are all non-English language, which mean “science.” The keyword “ouverte” and “abierta” all means “open” in English.

How to evaluate the performance of open science is always an important research problem. The “open evaluation” sub-theme began in 2012, a year after the open science starting year. The keyword “altmetrics” is non-traditional bibliometrics performed as an alternative or complement to more traditional citation impact metrics, such as impact factor and h-index.

2017 and 2018 witnessed the growth of blockchain. The keyword “linkedresearch” is a movement to encourage researcher to publish in a self-controlled and accessible way so that the free linkage to other research and reuse could be applied. “crypto” and “bitcoin” are all digital currency based on blockchain. These keywords show that blockchain technologies are applied to both open science research platform or tools, as well as to the finance.

The sub-theme “discipline” is the branches of learning that closely related to open science. The keywords show that “bioinformatics,”“climate,”“digital health,”“space,” etc., are all interconnected with open science. Instead of these traditional disciplines, open science could also facilitate novel subbranch of learning, such as “open sociology.”

Results Analysis on Scopus Data

Annual Hot Spots Analysis

To obtain the annual hot spots in scholarly research, we also applied temperature metrics to Scopus dataset. We extracted the keywords, title, and abstract of the research publications to get the hot spots. The results were shown in Table 8. In Table 8, top 10 keywords for different temperature values (T = 0.1, 0.3, 1.0) were described. It could be observed that the smaller temperature value will give more special keywords, while the high-temperature words tend to be fixed words and changed slowly over years. To get the annual hot spots of the Scopus data, the top 10 keywords were selected when T = 0.3.

Table 8.

Keywords for Different Temperature Values.

year	Temperature	Top 10 keywords for brief
2011	T = 0.1	cpass, authorization, ldap, svopme, saber, xquery, bioclipse, cybergis, teragrid, xpath
	T = 0.3	cpass, authorization, grid, ldap, svopme, saber, privilege, founder, xquery, bioclipse
	T = 1.0	grid, system, data, site, environment, resource, cpa, research, information, authorization
2012	T = 0.1	keeneland, percolation, newsletter, contour, benkler, kid, moratorium, talkoot, n3phele, hippel
	T = 0.3	data, keeneland, percolation, hadoop, contour, newsletter, mapreduce, benkler, kid, moratorium
	T = 1.0	data, research, grid, community, software, resource, analysis, user, management, tool
2013	T = 0.1	onama, agris, diybio, delegation, euchinahealthcloud, condo, fmridc, collage, milxxplore, genocon
	T = 0.3	onama, agris, delegation, data, diybio, inquiry, store, euchinahealthcloud, condo, fmridc
	T = 1.0	data, research, system, information, application, knowledge, computing, community, study, access
2014	T = 0.1	bronchiectasis, dihydrospiro, piperidine, wrangler, elastography, sandhills, aptamers, overshadow, dinaa, cyberleninka
	T = 0.3	bronchiectasis, dihydrospiro, piperidine, wrangler, mediation, phantom, research, street, elastography, sandhills
	T = 1.0	research, system, information, software, resource, management, performance, infrastructure, result, grid
2015	T = 0.1	steel, coalescent, sonification, pypedia, expedition, microenvironment, colon, gcips, irgd, samt2
	T = 0.3	data, tumor, coalescent, expedition, sonification, cancer, metastasis, microenvironment, pypedia, colon
	T = 1.0	data, cell, experiment, cancer, information, reproducibility, access, tumor, system, project
2016	T = 0.1	gleon, distractors, distractor, vocalisation, kademlia, braf, emerald, craf, cam, exosomes
	T = 0.3	gleon, distractors, vocalisation, kademlia, braf, adopter, tcga, bosc, tenant, craf
	T = 1.0	information, analysis, study, source, experiment, access, reproducibility, result, model, network
2017	T = 0.1	firmware, fer2016, d’anthropologie, revue, dysplasia, lateralisation, hiding, grace, chembench, xrootd
	T = 0.3	data, firmware, fer2016, lateralisation, d’anthropologie, information, laterality, revue, dysplasia, system
	T = 1.0	data, information, study, system, access, analysis, software, network, management, resource
2018	T = 0.1	secretase, dysbindin, cerebellum, rhoa, rotenone, bace, fingolimod, fyco1, ccl5, taboo
	T = 0.3	research, protein, secretase, dysbindin, cerebellum, cell, rhoa, rotenone, bace, mouse
	T = 1.0	research, protein, cell, study, analysis, information, article, system, practice, mouse
2019	T = 0.1	rrid, lactate, amphetamine, dispersal, meth, sajip, efflux, hspb8, testosterone, soccer
	T = 0.3	badge, receptor, study, article, testosterone, rat, practice, neuron, perfusion, rrid
	T = 1.0	study, article, practice, analysis, badge, information, receptor, cell, result, model
2020	T = 0.1	radiography, happiness, bottleneck, hemorrhoid, cataract, cancellation, betacoronavirus, beef, scutellaria, lane
	T = 0.3	study, analysis, health, review, practice, article, betacoronavirus, radiography, happiness, bottleneck
	T = 1.0	study, analysis, health, review, practice, article, result, information, method, system
2021	T = 0.1	ppds, pocus, marsh, neon, feminist, hypermobility, powder, retinol, heteroscedasticity, disintegration
	T = 0.3	study, review, analysis, health, ppds, article, practice, pocus, covid-19, method
	T = 1.0	study, review, analysis, health, article, practice, method, result, researcher, quality
2022	T = 0.1	cable, inhaler, pride, heliophysics, mtfs, oximetry, zikv, eszopiclone, splint, pigmentation
	T = 0.3	study, review, health, analysis, practice, method, intervention, article, trial, pride
	T = 1.0	study, analysis, review, health, practice, article, method, result, model, framework

We observed the corresponding research papers and gave explanations to some special representative keywords every year. From these explanations, it could be observed that the annual hot spots of the research papers are mainly technical terms in research or neologism about novel technologies, especially in biology, computer science, medicine, etc. For example, the keywords “grid computing” (2011), “hadoop” (2012), and “mapreduce” (2012) are all technology innovations and the precursors of the new technique “cloud computing” in 2014. It demonstrates that the annual research hotspots might predicate new technologies in the industry.

2011

Cpass: Comparison of Protein Active-Site Structures.

Ldap: Lightweight Directory Access Protocol.

Grid: grid computing.

Svopme: Scalable Virtual Organization privileges management environment.

Saber: Sounding of the Atmosphere using Broadband Emission Radiometry.

Xquery: XML query.

Bioclipse: an open source workbench for chemo- and bioinformatics.

2012

Keeneland: Keeneland Initial Delivery System (KIDS).

Percolation: percolation problem.

Hadoop: an open-source data processing framework including mapreduce model.

Mapreduce: fault-tolerant and scalable distributed data processing model and execution environment.

2013

Onama E-Onama: Mobile high performance computing for engineering research.

AGRIS: International Information System for the Agricultural Sciences and Technology.

Diybio: do it yourself biology.

Euchinahealthcloud: The EUChinaHealthCloud proposal contributes to the aims of the Research Infrastructures part of the EU Seventh Framework Programme (FP7) by promoting cloud computing between the European and the China.

Condo: Condominium cluster.

Fmridc: the FMRI Data Center.

2014

Bronchiectasis: disease.

Dihydrospiro: a kind of chemical production.

Piperidine: a kind of chemical production.

Wrangler: the Texas Advanced Computing Center presents Wrangler.

Phantom: phantom study.

Street: Openstreetmap, an open map tool.

Elastography: imaging the elastic properties of soft tissues.

Sandhills: distributed execution platforms for Scientific workflows.

2015

Coalescent: a chemical product for coalescing.

Metastasis: cancer metastasis.

Pypedia: using the wiki paradigm as crowd sourcing environment for bioinformatics protocols.

Colon: The section of the large intestine extending from the cecum to the rectum.

2016

Gleon: Global Lake Ecological Observatory Network.

Vocalisation: Ultrasonic vocalisation.

Kademlia: A kind of Distributed Hash Table.

Braf: a kind of gene for cancer biology.

Tcga: The Cancer Genome Atlas.

Bosc: host site.

Craf: a kind of gene for cancer biology.

2017

Firmware: a kind of hardware.

fer2016: Finkel et al. (2017), a research publication.

Lateralisation: language lateralisation.

d’anthropologie: anthropology in French.

Laterality: language laterality.

Dysplasia: maldevelopment.

2018

Secretase: gamma secretase.

Dysbindin: lack of bond.

Cerebellum: the part of the brain at the back of the head that controls the activity of the muscles.

Rhoa: a kind of gene.

Rotenone: a broad-spectrum insecticide.

Bace: beta-site APP-cleaving enzyme.

2019

Badge: open science badge.

Testosterone: a kind of chemical substance produced in the body.

Perfusion: cerebral perfusion.

Rrid: research resource identifiers.

2020

Betacoronavirus: a kind of coronavirus.

Radiography: the process or job of taking X-ray photographs.

2021

ppds: purified protein derivative-standard.

Pocus: point-of-care ultrasound.

2022

Trial: biology experiments.

In 2013, the keyword “diybio” is about citizen science. This is consistent with the time of the open science community booming mentioned in annual hot spots in tweet dataset. Also, open science tool (e.g., open street map) in 2014 was also mentioned in 2014 hot spots in tweet dataset. It is evident that the annual hot spots of open science from both scholarly research and the public opinions are consistent. From the annual hot spot of the academic, it could be noted that with the development of open science, social science becomes one of the research hot spots besides the technologies. For example, there are research publications about language laterality and anthropology (d’anthropologie) in 2017, as well as psychology such as pride (2020) and happiness (2022). The COVID-19 pandemic are mentioned in 2020 (betacorahavirus) and 2021 (COVID-19).

Keywords Co-occurrence Analysis

In our study, VOSviewer software application was employed for creating a network visualization map. We adopted VOSviewer application to create the co-occurrence of keywords, which analyze and count the paired data and similarity of the document (Ariza-Garzon et al., 2021). As shown in Figure 9, the network is made up of 1,682 keywords and six clusters, which have occurred at least five times in the data set.

Figure 9.

The visualization of keyword co-occurrence.

As shown in Figure 9, the largest cluster, represented by the red color, is made up of 665 keywords. The most prominent keywords (five for brief) in this cluster are open science, information management, software, open systems, publishing. These words shows the important aspects of the research on open science practice which describe the trends and focus on open science for the researcher in academic (Beamer, 2019; A. Bennett et al., 2022; Kusters & Klages, 2019). This cluster also indicated the evolution of research in the domain of information management, software, data handling, publish, etc. (Figueiredo et al., 2022; Mosconi et al., 2022).

The second largest cluster, which is in green, is based on 431 keywords. The top 5 prominent keywords in this cluster are human, female, male, systematic review, article. This cluster signifies the importance of the application and practice of open science in the research fields related to human being (O’Connor et al., 2022; Riches & Jackson, 2018; Smart et al., 2022). The application technology included systematic review, review literature as topic, etc. (A. Bennett et al., 2022; Carbine et al., 2019).

The third largest cluster (blue) consists of 244 keywords. The top 5 prominent keywords in this cluster are controlled study, priority journal, nonhuman, metabolism, and drug. These words shows the important aspect of open science in animal experiment, controlled study, drug, etc. (Barreda-Manso et al., 2021; Marsden et al., 2022).

The fourth largest cluster (grass green) is made up of 122 keywords. The top 5 prominent keywords in this cluster are ethics, information processing, procedures, biomedical research, and personnel. This cluster described the ethics consideration and the information processing and dissemination method for researchers in biology adopted open science practice (Bourgeat et al., 2013; Thumbeck et al., 2021).

The fifth largest cluster (purple) is based on 112 keywords. Reproducibility, human experiment, replication, preregistration, and meta research are top 5 keywords in this cluster. These keywords specify the nature of open science and the basic methods for researchers to practice open science, such as meta analysis, meta research, etc. (Draborg et al., 2022; Jackevicius et al., 2019).

The smallest cluster (sky blue) consist of 108 keywords. The top 5 keywords in this cluster are brain, neuro-science, diagnostic imaging, image analysis, and algorithm. This cluster described the open science practice on brain/neuro science and the implementation method, such as image processing, diagnostic imaging, algorithm, etc. (W. Bennett et al., 2018; Lefaivre et al., 2019).

Topics Analysis

We applied topic modeling on the Scopus dataset and got 10 topics. The topics and their proportions, as well as the top 5 keywords were shown in Table 9.

Table 9.

Topics and Keyword of the Scopus Data.

Topic (proportion in percent)	Top 5 keywords
Cell, protein and gene (5.1)	cell; protein; mouse; receptor; gene
Psychiatric illnesses and symptoms (1.2)	anxiety; addiction; rcts; alcohol; disorder
Biology (0.7)	plant; rat; radiomics; fmri; biomarkers
Open science practice (68.7)	access; practice; information; policy; management
Human subject or respondent (0.2)	firm; campus; meditation; scheduling; parent
Open data technology (1.1)	reuse; heritage; blockchain; identity; speech
Open science infrastructures (6.9)	grid; cloud; computing; system; infrastructure
Image process and brain science (8.2)	image; imaging; brain; learning; language
Human health (7.1)	health; review; trial; care; intervention
Infant and pregnancy (0.7)	infant; retraction; competence; pregnancy; birth

It could be observed in Table 9 that the largest cluster was topic “open science practice” (68.7%). This topic was mainly about open access, open science policy, open science practice, etc. (Hoces De La Guardia et al., 2021; Lefaivre et al., 2019). The second largest topic described image process and brain science (8.2%) (W. Bennett et al., 2018; Lefaivre et al., 2019). The third largest was “human health” topic, whose keywords included health, care, trial, etc. (7.1%) (Coro, 2020; O’Brien et al., 2021). The smallest cluster was about human subject or respondent (0.2%) (De Crescenzo et al., 2022; Klein, 2022). The human subject/respondent might come from firm or campus, and the experimental methods could be meditation, mindfulness, etc. (Bokk & Forster, 2022).

As Table 9 shows, the topic modeling results were consistent with the clusters of the keywords co-occurrence analysis: open science practice was the main focus for researchers towards open science. At the same time, open science infrastructures (6.9%) and open data technology (1.1%) were also mentioned in the academic papers. Besides the discussion on the components of open science, researchers were interested in open science practice in various disciplines, especially brain science (8.2%), human health (7.1%), gene (5.1%), biology (0.7%), psychiatric illnesses (1.2%), human reproduction (0.7%), etc.

Comparison Between Two Kinds of Viewpoints

When make comparison analysis between the tweet dataset and the Scopus dataset, we could find several noteworthy points. At first, from topic modeling, it could be observed that the topics about the open science knowledge and infrastructure attracted the most attention from both the public (55%) and the academic (76.7%). From these topics, open access was the hottest one. There were 24.5% posts from tweet data belonged to open access and scientific publications, while about 68% publications from Scopus data mentioned open access and open policy. Also, open science infrastructure was mentioned from both datasets, that is, 6% from the public and 6.9% from the researchers. Then, the topics about open data was more popular on social media (12.5%) then in academic (1.1%).

Secondly, the technologies of computer science have fostered open science practice from every aspect. From the analysis both on tweet data and Scopus data, it could be observed that the information technologies exerted important influence on the development of open science. Blockchain, for example, was a innovative technology for both open data management and the open science workflow (Tschirner et al., 2021; Yun-Chi et al., 2022). Cloud and grid computing, on the other hand, were the basic block for constructing the open science infrastructure (Fisser et al., 2020; Laurin et al., 2022). Also, the computer and information technologies were combined with other disciplines in open science practice. In brain science, for example, imagine processing and analysis were mentioned in open science practice (Gold, 2016; Mosser et al., 2021).

Thirdly, the topics about open engagement of societal actors were discussed more frequently on Twitter than in academic. Many citizen science projects were advertised and organized through Twitter, such as pdftribute, openoxford, phdvoice, etc. Hence, these projects or activities were frequently discussed on Twitter. There were also some studies on citizen science or crowd sourcing in Scopus data (Heigl et al., 2020; Sarpong et al., 2020), but it was not the burning issue for the researchers. The underlying idea might be that the public is more interested in open engagement than the academic since the public is the main participants for these projects. The public engagement of open science on social media reflected the urgent needs of the public to get benefit from the open science practice.

At last, by comparing the annual hot spots of the Twitter and the scholarly research, it could be noticed that the public on Twitter is more concerned about the social events and activities in open science, while the scholarly researchers focus on novel technologies and detailed research aspects. This difference between the annual hot spots on Twitter and in research publications further confirm the prior research results that the public pay more attention to the macroscopic aspects, while academia focuses on the more insightful and detailed spots (Q. Zhou & Zhang, 2021). In addition, technology hotspots might appear much earlier in scholarly research than on Twitter. For example, “Rrid” is mentioned in annual hot spots of research publications in 2019 and tweet dataset in 2022. Also, cloud computing appears in Scopus dataset in 2013 as “Euchinahealthcloud” and in tweet dataset as “Cambiare” in 2015.

Conclusion

Open science aims to make scientific knowledge openly available, accessible and reusable; to increase scientific collaborations and sharing of information; and to open the processes of scientific knowledge creation, evaluation and communication to societal actors. In this paper, a novel research framework has been proposed to disentangle the interest on Twitter and scholarly research towards open science, tracking the hot spots both from the public opinion and the academic view. For public opinion mining, after collected the tweets on open science, temperature metrics and topic modelling techniques were adopted to analyze the annual hot spots and the topics and themes. As to the academic view, VOSviewer application was employed for the visualization of the keyword co-occurrence. Also, LDA method was implemented for topic analysis. The comparison analysis between the two viewpoints towards open science were conducted. The results show that there were different needs for open science from the public and researchers though they are both interested in open science knowledge and infrastructure.

To address the first research question in this paper, themes/sub-themes, as well as the annual hot spots were mined from the public opinion on open science. Though open scientific knowledge is the most concerned theme for the public on social media, some breaking technologies or current events also attract attention. On one hand, the innovative technologies, such as cloud computing and blockchain facilitated the development of open science so that there were more novel open science research or practice. On the other hand, the public health emergencies, such as the COVID-19, promoted the public needs for open science and encouraged collaboration and open innovation.

In answer to the second research question, through the analysis of the Scopus data, the annual hot spots, keywords co-occurrence, and the topics modeling of open science could provide the insight for the development of open science in academic. Besides exploring the open science practice itself, the researchers are all interested in open science practice in their own fields, especially those in health, biomedicine, computer science, human and social, etc. The principles advocated by open science, such as reproducibility, open access, transparency, could help open cooperation in the researchers and thus lead to open innovation in academic.

As to the last research question, the results of the comparison analysis reflect that there were different needs towards open science from the public and the academic. The public and the researchers might have the similar interests towards open science, including open access, open data, open education, open infrastructure, etc. Based on these basic practice of open science, the public were more interested in open engagement of societal actors since there were intensive discussion on citizen science projects on Twitter. The academia were concerned more to make the research innovations.

For the future work, we planned to embedded semantics of words to get more precise topic analysis. At the same time, the research framework to get broader vision by comparing the viewpoints from the public and the academic could help policy-decision in other measuring problems. Hence, more bibliometrics analysis could be adopted in future experiments.

Limitation

Our research is limited in some aspects, and we call for future studies to extend our understanding of open science. First, the public opinion mining is applied on the tweet dataset collected from Twitter. Though Twitter is influential, it might be limitedly accessible in some countries. Therefore, this could lead to bias in this study. Future studies might consider integrating different social media platforms in the world to investigate the connections between the general public as well as professionals to make the results more robust. Second, LDA modeling was used in this study to identify topics, but LDA itself had limitations. LDA modeling could only cluster the words of the same topic through the probabilities of these word. The semantic similarities of the words are not adopted in LDA. As a result, our topics might be biased. In future studies, other method of topic modeling could be considered to adopt the semantics similarities of the words to get more precise classification.

Footnotes

Acknowledgements

Acknowledgements for Anonymous Reviewers. This work is supported by the National Social Science Foundation Project (No. 21BTQ030).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work is supported by the National Social Science Foundation Project (No. 21BTQ030).

ORCID iD

Wei Yu

Data Availability Statement

The article’s supporting data and digital research materials can be accessed on https://github.com/cjp360/openscience/blob/main/openscience.7z.001 and .

References

Al-Raeei

Al-Jabban

M. O.

Azmeh , et al. (2023). Bibliometrics study of the cancer research in the faculties of medicine and dentistry during Syrian crisis. Oral Oncology Reports, 8, 100105.

Alexopoulos

Zuiderwijk

Charapabidis

Loukis

Janssen

(2014). Designing a second generation of open data platforms: Integrating open data and social media [Conference session]. Electronic Government: 13th IFIP WG 8.5 International Conference, EGOV 2014, Dublin, Ireland, September 1–3, 2014. Proceedings 13 (pp. 230-241). Springer.

Ariza-Garzon

M. J.

Segovia-Vargas

M. J.

Arroyo

(2021). Risk return modelling in the p2p lending market: Trends, gaps, recommendations and future directions. Electronic Commerce Research and Applications, 49, 101079.

Barreda-Manso

M. A.

Nieto-Díaz

Soto

Muñoz-Galdeano

Reigada

Maza

R. M.

(2021). In silico and in vitro analyses validate human micrornas targeting the sars-cov-2 3′-utr. International Journal of Molecular Sciences, 22(11), 6094.

Bartling

Friesike

(2014). Opening science. Springer.

Bashar

Nyagadza

Ligaraba

, et al. (2023). The influence of Covid-19 on consumer behaviour: a bibliometric review analysis and text mining. Arab Gulf Journal of Scientific Research.

Beamer

J. E.

(2019). Digital libraries for open science: Using a socio-technical interaction network approach [Conference session]. The Proceedings of IRCDL 2019.

Bennett

Beck

Shaver

Grad

LeBlanc

Limburg

Gray

Abou-Setta

Klarenbach

Persaud

Thériault

(2022). Screening for prostate cancer: Protocol for updating multiple systematic reviews to inform a Canadian Task Force on Preventive Health Care guideline update. Systematic Reviews, 11(1), 230.

Bennett

Smith

Jarosz

Nolan

Bosch

(2018). Reengineering workflow for curation of DICOM datasets. Journal of Digital Imaging, 31(6), 783–791.

10.

Besancon

Peiffer-Smadja

Segalas

Jiang

Masuzzo

Smout

Billy

Deforet

Leyrat

(2021). Open science saves lives: Lessons from the COVID-19 pandemic. BMC Medical Research Methodology, 21(1), 117.

11.

Blair

S. J.

Mulvenna

M. D.

(2020). Aggregated topic models for increasing social media topic coherence. Applied Intelligence, 50(1), 138–156.

12.

Blei

D. M.

A. Y.

Jordan

M. I.

(2003). Latent Dirichlet allocation. Journal of Machine Learning Research, 3, 993–1022.

13.

Boby

Fearon

Ferla

Filep

Koekemoer

Robinson

M. C.

COVID Moonshot Consortium Chodera

J. D.

Lee

A. A.

London

von Delft

(2023). Open science discovery of potent noncovalent SARS-CoV-2 main protease inhibitors. Science, 382(6671), eabo7201.

14.

Bokk

Forster

(2022). The effect of a short mindfulness meditation on somatosensory attention. Mindfulness, 13(8), 2022–2030.

15.

Bourgeat

Dore

Villemagne

V. L.

Rowe

C. C.

Salvado

Fripp

(2013). MilxXplore: A web-based system to explore large imaging datasets. Journal of the American Medical Informatics Association, 20(6), 1046–1052.

16.

Brembs

Lenardic

Murray-Rust

Chan

Irawan

D. E.

(2023). Mastodon over mammon: Towards publicly owned scholarly knowledge. Royal Society Open Science, 10(7), 230207.

17.

Brown

T. B.

Mann

Ryder

Subbiah

Kaplan

J. D.

Dhariwal

Neelakantan

Shyam

Sastry

Askell

Agarwal

(2020). Language models are few-shot learners [Conference session]. 34th Conference on Neural Information Processing Systems (NeurIPS 2020).

18.

Buchi

(2017). Microblogging as an extension of science reporting. Public Understanding of Science, 26(8), 953–968.

19.

Carbine

K. A.

Lindsey

H. M.

Rodeback

R. E.

Larson

(2019). Quantifying evidential value and selective reporting in recent and 10-year past psychophysiological literature: A pre-registered P-curve analysis. International Journal of Psychophysiology, 142, 33–49.

20.

Chandrasekaran

Mehta

Valkunde

Moustakas

(2020). Topics, trends, and sentiments of tweets about the COVID-19 pandemic: Temporal infoveillance study. Journal of Medical Internet Research, 22(10), e22624

21.

Cheplygina

Hermans

Albers

Bielczyk

Smeets

(2020). Ten simple rules for getting started on Twitter as a scientist. PLoS Computational Biology, 16(2), e1007513.

22.

Choi

(2023). Revisiting and updating the state of library open source software research. The Electronic Library, 41(1), 137–151.

23.

Coro

(2020). A global-scale ecological niche model to predict SARS-CoV-2 coronavirus infection rate. Ecological Modelling, 431, 109187.

24.

De Crescenzo

D’Alò

G. L.

Ostinelli

E. G.

Ciabattini

Di Franco

Watanabe

Kurtulmus

Tomlinson

Mitrova

Foti

Del Giovane

(2022). Comparative effects of pharmacological interventions for the acute and long-term management of insomnia disorder in adults: A systematic review and network meta-analysis. The Lancet, 400(10347), 170–184.

25.

Dordevic

B. S.

Jovanovic

S. P.

Timcenko

V. V.

(2014). Cloud computing in Amazon and Microsoft Azure platforms: Performance and service comparison [Conference session]. The Proceedings of TELFOR 2014.

26.

Draborg

Andreasen

Norgaard

Juhl

C. B.

Yost

Brunnhuber

Robinson

K. A.

Lund

(2022). Systematic reviews are rarely used to contextualise new results—A systematic review and meta-analysis of meta-research studies. Systematic Reviews, 11(1), 189.

27.

Dursun

Yildiz

(2022). New media and the public sphere: Perspectives of communication academics. Connectist-Istanbul University Journal of Communication Sciences, 62, 1–32.

28.

Fang

Costas

Wouters

(2022). User engagement with scholarly tweets of scientific papers: A large-scale and cross-disciplinary analysis. Scientometrics, 127, 4532–4546.

29.

Figueiredo

Scherer

Cabral

J. S.

(2022). A simple kit to use computational notebooks for more openness, reproducibility, and productivity in research. PLoS Computational Biology, 18(9), e1010356.

30.

Finkel

Eastwick

Reis

(2017). Replicability and other features of a high-quality science: Toward a balanced and empirical approach. Journal of Personality and Social Psychology, 113(2), 244–253.

31.

Fisser

Ipach

Timm-Giel

(2020). Evaluation of LTE based communication for fast state estimation in low voltage grids [Conference session]. The Proceedings of the 2020 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids, SmartGridComm 2020.

32.

Gold

E. R.

(2016). Accelerating translational research through open science: The neuro experiment. PLoS Biology, 14(12), e2001259.

33.

Heigl

Kieslinger

Paul

K. T.

(2020). Co-creating and implementing quality criteria for citizen science. Citizen Science: Theory and Practice, 5(1), 24.

34.

Hoces De La Guardia

Grant

Miguel

(2021). A framework for open policy analysis. Science and Public Policy, 48(2), 154–163.

35.

Homolak

Kodvanj

Virag

(2020). Preliminary analysis of COVID-19 academic information patterns: A call for open science in the times of closed borders. Scientometrics, 124, 2687-2701.

36.

Jackevicius

C. A.

D. T.

(2019). Submissions from the SPRINT Data Analysis Challenge on clinical risk prediction: A cross-sectional evaluation. BMJ Open, 9(3), e025936.

37.

Jia

J. L.

Nguyen

Mills

D. E.

Polin

D. J.

Sarin

K. Y.

(2020). Comparing online engagement and academic impact of dermatology research: An altmetric attention score and PlumX metrics analysis. Journal of the American Academy of Dermatology, 83(2), 648–650.

38.

Karami

Lundy

Webb

Dwivedi

Y. K.

(2020). Twitter and research: A systematic literature review through text mining. IEEE Access, 8, 67698–67717.

39.

Klein

Fairweather

A. K.

Lawn

(2022). The impact of educational interventions on modifying health practitioners’ attitudes and practice in treating people with borderline personality disorder: An integrative review. Systematic Reviews, 11(1), 108.

40.

Kung

Shiri

(2024). Text mining applications to support health library practice: A case study on marijuana legalization Twitter analytics. Health Information & Libraries Journal, 41(1), 53–63.

41.

Kusters

Klages

(2019). Fostering open science at Fraunhofer. Procedia Computer Science, 146, 39–52.

42.

Laurin

G. V.

Francini

Penna

(2022). SnowWarp: An open science and open data tool for daily monitoring of snow dynamics. Environmental Modelling and Software, 156, 105477.

43.

Lefaivre

Behan

Vaccarino

Evans

Dharsee

Gee

Dafnas

Mikkelsen

Theriault

(2019). Big data needs big governance: Best practices from brain-CODE, the Ontario-brain institute’s neuroinformatics platform. Frontiers in Genetics, 10, 191.

44.

Lefebvre

Spruit

(2023). Laboratory forensics for open science readiness: An investigative approach to research data management. Information Systems Frontiers, 25, 381–399.

45.

Luhmann

Burghardt

(2022), Digital humanities-a discipline in its own right? An analysis of the role and position of digital humanities in the academic landscape. Journal of the Association for Information Science and Technology, 73, 148–171.

46.

Marsden

Kelleher

Hoare

Hughes

Bisla

Cape

Cowden

Day

Dewhurst

Evans

Hearn

(2022). Extended-release pharmacotherapy for opioid use disorder (EXPO): Protocol for an open-label randomised controlled trial of the effectiveness and cost-effectiveness of injectable buprenorphine versus sublingual tablet buprenorphine and oral liquid methadone. Trials, 23(1), 697.

47.

Martin

Johnson

(2015). More efficient topic modelling through a noun only approach [Conference session]. Proceedings of Australasian Language Technology Association Workshop, ALTA 2015.

48.

Molldrem

Hussain

Smith

A. K.

(2021). Open science, COVID-19, and the news: Exploring controversies in the circulation of early SARS-CoV-2 genomic epidemiology research. Global Public Health, 16(8/9), 1468–1481.

49.

Mosconi

Randall

Karasti

(2022). Designing a data story: A storytelling approach to curation, sharing and data reuse in support of ethnographically-driven research [Conference session]. The Proceedings of the ACM on Human-Computer Interaction.

50.

Mosser

C.-A.

Haqqee

Nieto-Posadas

(2021). The McGill-Mouse-Miniscope platform: A standardized approach for high-throughput imaging of neuronal dynamics during behavior. Genes, Brain and Behavior, 20(1), e12686.

51.

Nosek

Alter

Banks

G. C.

et al. (2015). Promoting an open research culture. Science, 348(6242), 1422–1425.

52.

O’Brien

M. W.

Petterson

J. L.

Kimmerly

D. S.

(2021). An open-source program to analyze spontaneous sympathetic neurohemodynamic transduction. Journal of Neurophysiology, 125(3), 972–976.

53.

O’Connor

Mandino

Shen

Horien

Herman

Hyder

Crair

Papademetris

Lake

E. M. R.

Constable

R. T.

(2022). Functional network properties derived from wide-field calcium imaging differ with wakefulness and across cell type. NeuroImage, 264, 119735.

54.

Orduna-Malea

Bautista-Puig

(2024). Research assessment under debate: Disentangling the interest around the DORA declaration on Twitter. Scientometrics, 129, 537–559.

55.

Ramachandran

Bugbee

Murphy

(2021). From open data to open science. Earth and Space Science, 8(5), e2020EA001562.

56.

Riches

Jackson

(2018). Individual differences in syntactic ability and construction learning: An exploration of the relationship. Language Learning, 68(4), 973–1000.

57.

Rodriguez-Pomeda

Casani

Serrano-López

A. E.

(2023). Reflections on the diffusion of management and organization research in the context of open science in Europe. European Management Journal, 41, 664–672.

58.

Sadiq

M. T.

Yadav

A. K.

(2022). Discovering the open access movement on Twitter: An exploratory study. Journal of Indian Library Association, 57(1), 67–77.

59.

Sarpong

Ofosu

Botchie

(2020). Do-it-yourself (DiY) science: The proliferation, relevance and concerns. Technological Forecasting and Social Change, 158, 120127.

60.

Sedighi

(2023). Altmetrics analysis of selected articles in the field of social sciences. Global Knowledge Memory and Communication, 72(4/5), 452–463.

61.

Smart

K. M.

Hinwood

N. S.

Dunlevy

Doody

C. M.

Blake

Fullen

B. M.

Le Roux

C. W.

O’Connell

Gilsenan

Finucane

F. M.

O’Donoghue

(2022). Multidimensional pain profiling in people living with obesity and attending weight management services: A protocol for a longitudinal cohort study. BMJ Open, 12(12), e065188.

62.

Sotudeh

(2023). How social are open-access debates: A follow-up study of tweeters’ sentiments. Online Information Review, 48(1), 159–186.

63.

Sotudeh

Saber

Ghanbari Aloni

Mirzabeigi

Khunjush

(2022). A longitudinal study of the evolution of opinions about open access and its main features: A Twitter sentiment analysis. Scientomentrics, 127(10), 5587–5611.

64.

Stewart

Walker

(2018). Build it and they will come? Patron engagement via Twetter at historically black college and university libraries. Journal of Academic Librarianship, 44(1), 118–124.

65.

Steyvers

Griffiths

(2007). Probabilistic topic models. In Landauer

T. K.

McNamara

D. S.

Dennis

(Eds.), Handbook of latent semantic analysis. Lawrence Erlbaum Associates.

66.

Taylor

(2023). Slow, slow, quick, quick, slow: Five altmetric sources observed over a decade show evolving trends, by research age, attention source maturity and open access status. Scientometrics, 128, 2175–2200.

67.

Thorpe

Gregory

(2010). Producing the post-Fordist public: The political economy of public engagement with science. Science as Culture, 19(3), 273–301.

68.

Thumbeck

S.-M.

Schmid

Chesneau

Domahs

(2021). Efficacy of a strategy-based intervention on text-level reading comprehension in persons with aphasia: A study protocol for a repeated measures study. BMJ Open, 11(7), e048126.

69.

Troussas

Krouska

Virvou

(2019). Trends on sentiment analysis over social networks: Pre-processing ramifications, stand-alone classifiers and ensemble averaging. In Tsihrintzis

Sotiropoulos

Jain

(Eds.), Machine learning paradigms, intelligent systems reference library (Vol. 149, pp. 161–186). Springer.

70.

Tschirner

Röper

Zeuch

(2021). Fostering open data using blockchain technology. Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, 378, 209–228.

71.

UNESCO. (2021). UNESCO recommendation on open science. Retrieved June 12, 2022, from https://en.unesco.org/science-sustainable-future/open-science/recommendation

72.

Voytek

(2017). Social media, open science, and data science are inetricably linked. Neuron, 96(6), 1219–1222.

73.

Walsh

J. A.

Cobb

P. J.

Fremery

, et al. (2022). Digital humanities in the iSchool. Journal of the Association for Information Science and Technology, 73, 188–203.

74.

Wang

Chen

(2023). Identifying interdisciplinary topics and their evolution based on BERTopic. Scientometrics, 2023, 1–26.

75.

Yun-Chi

Li-Fei

Hsu-Chun

(2022). “Prove It!”: A user-centric client for the blockchain-based research lifecycle transparency framework [Conference session]. The Proceedings of the Association for Information Science and Technology.

76.

Zhang

Gou

Fang

Sivertsen

Huang

(2023). Who tweets scientific publications? A large-scale study of tweeting audiences in all areas of research. Journal of the Association for Information Science and Technology, 74, 1485–1497.

77.

Zhang

Xie

Liang

(2024). How is public discussion as reflected in WeChat articles different from scholarly research in China? An empirical study of metaverse. Scientometrics, 129, 473–495.

78.

Zhou

Jing

(2020). Multidimensional mining of public opinion in emergency events. The Electronic Library, 38(3), 545–560.

79.

Zhou

Zhang

(2021). Breaking community boundary: Comparing academic and social communication preferences regarding global pandemics. Journal of Informetrics, 15(3), 101162.

80.

Zhou

Zhao

Bian

Haynos

A. F.

Zhang

(2020). Exploring eating disorder topics on Twitter: Machine learning approach. JMIR Medical Informatics, 8(10), e18273.

81.

Zong

Huang

(2023). Do open science badges work? Estimating the effects of open science badges on an article’s social media attention and research impacts. Scientometrics, 128, 3627–3648.

Open Science Under Debate: Disentangling the Interest on Twitter and Scholarly Research

Abstract

Keywords

Introduction

Data and Methodology

Research Design

Tweets Dataset

Scopus Dataset

Temperature Metrics

Topic Modelling and Aggregation

Assigning Themes in Public Opinions

Results Analysis on Tweets Data

Annual Hot Spots Analysis

2011 The Starting Year of Open Science: Some First-Time Discussion or Contest to Initiate Open Science

2012 Open Science Policy Establishment: Some Rules and Act Established as Basic Blocks of Open Science

2013 The Boom of Communities and Events on Open Science

2014 Tools and Services on Open Science

2015 the Beginning of the EU Cloud and Open Science Practice

2016 Practice on Open Science and Zika Virus

2017 Open Science Practice on Farming and Citizen Science

2018 Blockchain and Open Science

2019 Open Science Based on Blockchain and Online Course

2020 COVID-19

2021 More Open Science Practice, Especially on Health

2022 More Diverse Open Science Practice, Including Different Infrastructure, Discipline, or Gender

Themes and Prevailing Trends

Theme 1: Open Scientific Knowledge

Theme 2: Open Science Infrastructures

Theme 3: Open Engagement of Societal Actors

Theme 4: Others

Results Analysis on Scopus Data

Annual Hot Spots Analysis

2011

2012

2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

Keywords Co-occurrence Analysis

Topics Analysis

Comparison Between Two Kinds of Viewpoints

Conclusion

Limitation

Footnotes

Acknowledgements

Declaration of Conflicting Interests

Funding

ORCID iD

Data Availability Statement

References