Introducing NexSys research through the lens of publication abstracts

by Brian Boyle and Stefan Müller

As political and computational social scientists, at NexSys we are exploring how researchers, parties, politicians, interest groups, and legal documents discuss net-zero emissions policies. Using quantitative text analysis and supervised machine learning, we will define and map policies relating to the environment and sustainability, and provide recommendations for policymakers. This brief analysis forms part of our broader work within NexSys.

The authors

Keen to learn more about NexSys and unsure where to start? In this blog post, we use bibliometrics, the statistical analysis of publications, to introduce the NexSys team and their research. 

We were curious to find out how our team’s prior work relates to the core aims of the NexSys programme, and which issues relating to the strands of NexSys have been addressed in past abstracts of publications. Our results illustrate the depth and breadth of NexSys research.

Over 2,600 publications from ten different subject areas

We systematically collected information on previous academic publications from the NexSys team. The Elsevier Scopus database contains abstracts and citation information on over 85 million documents, across more than 25,000 peer-reviewed journals, books, and conference papers. Of the 74 NexSys team members listed on the project staff page [data collected in January 2023], 59 were present in the Scopus database. This covered all staff members with the exception of our PhD Students and non-academic members of the team.

While over half of NexSys researchers have an engineering background, the full team covers ten different subject areas, including architecture, computer science, economics, and social policy. The engineers themselves come from over seven sub-fields, including chemical, civil, electrical, marine, and mechanical engineering.

Searching the Scopus database by author returned 3,200 publications, 2,880 of which contained a valid digital object identifier (DOI), and relevant summary description text (e.g. article abstracts). The NexSys members’ publications were spread across a variety of formats, including 2,000 journal articles, 600 conference papers, as well as 160 books and book chapters.

Differences Across Disciplines

We provide descriptive analyses of publication abstracts using the quanteda R package (Benoit et al. 2018) for quantitative text analysis. The table below lists the number of abstracts from each subject area. We also report the abstracts’ most frequent terms and phrases, after removing punctuation characters, numbers, and so-called “stopwords” which appear in almost all scientific publications. 

The list underscores the depth of our research, but also shows that researchers from most disciplines have directly worked on one or more of the core issues of the NexSys programme.

Most Frequent Features in Publication Abstract by Subject Area

• Architecture (136 abstracts): nbs (62 mentions), performance (54), impact (48), ireland (43), building (43), design (36), monitoring (36), urban (35), ess (35), concrete (33)

• Business (127 abstracts): ireland (37 mentions), policy (35), consumers (34), energy (33), social (33), research (28), firms (28), policies (27), acceptability (26), demand (25)

• Chemical Engineering (176 abstracts): biofilm (138 mentions), biofilms (69), membrane (69), process (63), potential (55), system (54), high (53), oxygen (49), gas (48), treatment (46)

• Civil/Structural Engineering (409 abstracts): ireland (156 mentions), bridge (139), potential (134), cycling (125), vehicle (120), water (114), transport (110), new (107), performance (100), system (100)

• Computer Science (254 abstracts): system (173 mentions), performance (107), potential (78), systems (75), process (71), technology (71), students (69), learning (65), technologies (65), biomass (65)

• Economics (51 abstracts): carriers (38 mentions), policy (22), network (20), period (20), networks (18), airports (18), european (17), factors (16), regional (16), costs (16)

• Electrical Engineering (759 abstracts): system (325 mentions), control (220), power_system (166), power_systems (165), ieee (152), impact (147), frequency (145), power (141), network (131), proposes (129)

• Geography (61 abstracts): urban (105 mentions), cities (64), lcz (39), climate (37), city (28), surface (26), studies (24), wudapt (23), canyon (23), urban_areas (22)

• Marine Engineering (54 abstracts): wave (49 mentions), responses (45), response (37), wind (33), design (32), platform (32), compared (28), motion (27), wind_turbine (26), present (25)

• Mathematics (35 abstracts): ireland (22 mentions), flow (17), temperature (16), increase (15), future (15), applied (15), winter (14), observed (13), period (13), forecast (13)

• Mechanical Engineering (598 abstracts): building (166 mentions), performance (156), experimental (148), system (127), compared (125), potential (118), damage (112), flow (112), numerical (112), energy (107)

• Other/Non-academic (187 abstracts): system (68 mentions), adaptation (55), ieee (54), network (53), voltage (52), impact (49), methodology (48), control (46), load (45), power_systems (44)

• Politics and Social Policy (33 abstracts): housing (48 mentions), parties (24), voters (23), policy (20), social (17), electoral (15), support (14), problems (13), government (13), party (12)

The Focus on the Five Strands

NexSys consists of five strands: four hub strands (Water; Cities and Communities; Transport; Offshore Wind), and the Energy Systems core strand linking these four areas.

We explore how the NexSys team’s research fits into each strand (due to the overarching scope of the Energy Systems core strand, this was excluded from the current analysis). In order to classify the database of publication abstracts, we used a two-stage procedure. First, we selected ‘seed words’ that were narrowly and directly related to each strand (water; cities, city, community; transport, infrastructure; wind, offshore wind). We then checked whether or not an abstract contained none, one, or more than one of these keywords. 

With this initial simple classification, we moved on to so-called keyness analysis, a method through which frequent words can be identified (see, e.g., Bondi and Scott 2010; Zollinger 2022). Taking each strand in turn, we set the abstract texts that were identified as belonging to that strand based on the dictionary search as our target category, and all other abstract texts as the reference group. We then compared the relative frequency of features (this could be words or multi-word expressions) across each set of documents and identified words strongly associated with a specific category.

Words that occur very often in documents in our target category, but do not appear much in any of the reference documents, would produce a relatively high Chi2 value. Words that appear frequently in the reference documents, but not very often in the target documents would contain negative values.

From the keyness analysis, we took the ten most distinctive features for each strand and ran new dictionary searches using this expanded set of keywords to re-classify the publication abstracts, which are outlined below.

Strand Classification – Keyness Analysis 10 Most Distinct Features

• Water (525 abstracts): water; wastewater; wwtps; wastewater treatment; wastewater treatment plants; cod; treatment; reactor; leachate; underwater

• Transport (981 abstracts): transport; infrastructure; cycling; transportation; public transport; travel; car; cyclists; vehicles; avs

• Cities (1019 abstracts): capacity; electricity; cities; city; velocity; urban; community; communities; lcz; market

• Wind (529 abstracts): wind; wind turbine; wind power; wind generation; wind speed; wind turbines; wind farms; wind farm; wind power generation; turbine

After this second round of classification, we took each set of abstracts labelled under each strand and then ran a final keyness analysis. For abstracts falling into each of the four strands, we compared publications by engineers with researchers across all other subject areas. This allows us to explore how our team’s overall previous research aligns with the NexSys strands. The keyness analysis also helps us understand how the focus on each strand differs across research fields.

The results for each strand are displayed in the figures below. 

Broadly speaking, we observe clear differences in terms of both the language used across each of the four strands, as well as between engineering and other disciplines’ focus within each strand.

Engineering research tends to have a more focused scope, that is directly tied to concepts surrounding measurement, technology, and physical systems. The remaining research fields, meanwhile, tend to relate to a somewhat broader level of analysis, with the most distinct terms focusing on people-centred aspects of each strand. Examples include urban areas, specific locations and places, and the human impact of climate change. 

This very preliminary result highlights that NexSys researchers have focused on a combination of the technical questions and societal effects of these technologies. 

What we learned

Exploring the abstract texts from the NexSys team’s previous publications highlights the breadth of research experience and knowledge offered within the programme. It is clear that bringing together researchers from a broad range of subject areas is a key advantage of NexSys in its endeavour to develop technical, political, and social solutions to reach our net zero energy goals. This initial textual analysis shows how NexSys addresses the programme’s core objectives from various angles.

References

Benoit, Kenneth, Kohei Watanabe, Haiyan Wang, Paul Nulty, Adam Obeng, Stefan Müller, and Akitaka Matsuo. 2018. “Quanteda: An R Package for the Quantitative Analysis of Textual Data.” The Journal of Open Source Software 3 (30): 774. https://doi.org/10.21105/joss.00774

Bondi, Marina, and Mike Scott, eds. 2010. Keyness in Texts. Amsterdam: John Benjamins. https://doi.org/10.1075/scl.41

Zollinger, Delia. 2022. “Cleavage Identities in Voters’ Own Words: Harnessing Open-Ended Survey Responses.” American Journal of Political Science published ahead of print. https://doi.org/10.1111/ajps.12743

About NexSys

Next Generation Energy Systems (NexSys) is an all-island multidisciplinary research programme, involving nine different research institutions, alongside industry partners from across the energy sector. The programme’s key aims include tackling the challenges of energy decarbonisation, and developing evidence-based pathways for a just, net-zero energy system.