Sage Journals: Discover world-class research

Abstract

With the urgent call for sustainability and the rapid advancement of information technology, especially machine learning and artificial intelligence, the food industry faces unparalleled opportunities and challenges in optimizing resources and enhancing performance. By that call, we foresee a massive change, especially to the way research has been traditionally conducted, it is hypothesized for a surge of research and increasing adoption of innovative tools such as deep learning (DL) and optimization techniques by the research community and the field to address critical challenges in food systems. This study investigates the intersection of commercial, sustainability, and technology domains through advanced bibliometric analysis enhanced by deep document clustering, aims to uncover the multifaceted nature of these interconnected fields, addressing how emerging technologies and sustainability practices shape commercial strategies in the field, providing a panoramic perspective for reader about the dynamic development of the field. Utilizing a multi-tiered literature search strategy and employing deep document clustering techniques, the analysis reveals distinct thematic clusters—ranging from macroeconomic demand challenges to technical advancements in AI-driven agricultural practices and sustainable supply chain management—that underscore the deep connections among commercial, sustainability, and technological domains. The results not only validate the hypothesis of increasing convergence among these fields but also highlight the critical role of advanced methodologies in uncovering latent patterns and facilitating strategic and operational improvements. Moreover, the findings demonstrate that deep document clustering is a powerful tool for bibliometric analysis, offering nuanced insights that can be further refined in future studies. Overall, this research contributes to a holistic understanding of the field, emphasizing the need for continued interdisciplinary collaboration and innovation to drive sustainable development in a global business environment.

Keywords

sustainability food systems technology document clustering bibliometric analysis

Introduction

This paper aims to comprehensively map the evolving landscape of food systems,¹ specifically highlighting the impacts of various external pressures. We attempt to achieve this by leveraging advanced bibliometric techniques and multi-tiered literature search strategy^2,3 to uncover key trends and relationships.

The surge of call for sustainability and ESG

The intersection of food systems, sustainability, and technological innovation has long been a critical focus for both researchers and practitioners. Food systems encompass the entire spectrum of activities involved in producing, processing, distributing, and consuming food, thereby influencing environmental, economic, and social domains. The concept of sustainable food systems gained prominence. The foundational principles outlined in the 1987 BrundtHistorically, the study of food systems centered on production efficiency—as exemplified by the Green Revolution, which introduced high-yield crops and advanced agricultural techniques to meet the demands of growing populations⁴— often at the expense of environmental and social sustainability.^5–7 However, as global challenges such as climate change, resource depletion, and population growth emerged, land Report emphasized the need to meet current demands without compromising future generations,⁸ spurring research into life cycle assessments, circular economies, and ESG principles to mitigate the negative impacts of food systems.^9,10 For example, the initiatives have lead to establishment of program and guideline for life cycle assessments as a method in the fast growing food processing field as a method for environmental impact evaluations for the fast growing food processing field.¹¹ Circular economy has also gain the momentum to grow from a solution idea to feasibility and integration study to the food system, including academic, supply chain, economic, production and business model.^12–16 However, the urgency of climate change has make the call for sustainability within food systems evolve from general encouragement to a matter of serious regulatory concern. Initially framed within broad discussions of environmental responsibility and voluntary corporate social responsibility (CSR), sustainability is now increasingly embedded in food policy frameworks and legal mandates across various nations, expecting an enforcement in the future. For instance, the European Union’s Farm to Fork Strategy mandates sustainable practices from production to consumption, including EU code of conduct for responsible business and marketing practice, and related corporate governance framework and circular business models.¹⁷ While the United States introduced the Food Loss and Waste 2030 Champions initiative to engage industries in measurable waste reduction.¹⁸ Similarly, countries like China, Germany, France, and Japan, developing countries have imposed carbon accounting or food waste reporting requirements on large food companies.^19,20 These regulatory shifts signal a growing governmental commitment to enforceable sustainability in food systems.

The rise of new information processing technology

Another significant change comes from the rise of new technology, especially artificial intelligence (AI), machine learning (ML) and optimization techniques, represents a fundamental paradigm shift. These technologies are not merely tools but transformative infrastructures that permeate virtually every layer of the economic system. Their integration into the food system is particularly noteworthy by enabling sophisticated modeling and decision-making processes. Deep learning, for example, has been instrumental in improving food quality control, production and supply chain optimization, and precision agriculture, while both constrained and unconstrained optimization algorithms have enhanced resource allocation and waste reduction.^21–26

Impact to food production

With the rapid and high impact externally to the food system, the operational priorities of food systems have shifted dramatically. Modern food industries now integrate concerns for sustainability and ESG into their traditional production and efficiency models, resulting in a diversified, cross-disciplinary approach. This dual pressure—from sustainability demands and technological disruption—marks a structural inflection point in the evolution of food systems. The convergence of commercial food systems, sustainability, and technological innovation has thus created a fertile ground for interdisciplinary research. Modern food system operations are no longer confined to conventional production-centric metrics; they are now deeply intertwined with these external factors, drawing upon diverse domain knowledge from engineering, economics, environmental science, and data analytics. A deeper scholarly engagement is urgently needed to understand how these forces interact, conflict, or synergize to shape the future of global food production, distribution, and consumption. Unfortunately, by our knowledge, such a report is not available yet which clarifies the purpose of this paper.

Purpose of the study

By using multi-tiered search strategy, this study systematically examines the intersection degree of these fields by analyzing research productivity, thematic developments, and emerging trends. We hypothesize that the urgent need for sustainability, combined with rapid advances in information technology, will drive the application of transformative technological solutions to optimize resource utilization and overall performance within food systems. With the modern bibliometric tool, we carried out the bibliometric analysis attempt to test the hypothesis. Not only that, other than the conventional bibliometric analysis procedure, we have also adapted the newly developed, Deep Language Model Representation of Document Clustering with BERT and SBERT²⁷ into the analysis, in order to further uncover the emerging trends and latent themes. We compared the results with those of conventional methods. Overall, our research offers a comprehensive, panoramic perspective on the current research landscape and provides actionable insights for future innovation and interdisciplinary collaboration.

Bibliometric analysis

Bibliometric analysis is a technique emerged in the 60s as a powerful, data-driven approach to mapping scientific knowledge, uncovering research trends, and evaluating academic influence. Defined by Pritchard,²⁸ bibliometrics refers to the application of quantitative analysis and statistics to publications such as journal articles and their accompanying citation data. Unlike narrative or meta-analytical reviews, bibliometrics offers objectivity and reproducibility by leveraging structured databases (e.g., Scopus, Web of Science) to uncover co-authorship, co-citation, and keyword co-occurrence patterns across large datasets through correlation-based network analysis using structured metadata such as author names, references, and keywords, that form the foundation of bibliometric mapping, allowing for the visualization of relationships and thematic structures within a research domain.^29–32 Its advantages include transparency, scalability, and the ability to visualize complex intellectual structures and knowledge evolution. However, the shortcomings are also obvious, that the outcome are straight statistical relationships based on meta-data where semantic context is missing: they only give statistical links from data, without the full context.

Current tools: Bibliometrix and biblioshiny

One of the most widely adopted tools in modern bibliometric research is Bibliometrix, an R-based open-source software developed by.³³ It facilitates various analyses including performance metrics, science mapping, and thematic evolution through an accessible script-based interface. To make this functionality more approachable, Biblioshiny was introduced as its interactive web interface, enabling non-technical users to perform robust bibliometric analyses with minimal coding. These tools have gained strong adoption in management, sustainability, and engineering fields for their ease of use and analytical depth.^34,35 While these tools digitalize traditional bibliometric analysis, they inherently carry over its existing limitations.

Natural language processing (NLP) and its incorporation into bibliometric analysis

Natural Language Processing (NLP) is a subfield of artificial intelligence (AI) and linguistics focused on enabling machines to understand, interpret, generate, and respond to human language in a meaningful way. It integrates computer science, linguistics, and machine learning to bridge the gap between human communication and machine comprehension.^36,37 While traditional bibliometric techniques focus on citation-based metadata, NLP enables analysis at the full-text or abstract level, offering deeper semantic insights. NLP techniques such as topic modeling (e.g., LDA), named entity recognition, and sentiment analysis allow researchers to detect nuanced patterns within the content itself—beyond author keywords or citation links. This synergy of bibliometrics and NLP helps overcome limitations like synonymy, polysemy, and outdated keyword taxonomies in classical bibliometrics. Applications include automatic classification of documents, content-based trend analysis, and identification of research gaps, yielding new insights from many fields.^38–49

BERT and SBERT: Transforming text mining in bibliometrics

The introduction of transformer-based models like BERT (Bidirectional Encoder Representations from Transformers) has revolutionized NLP by enabling deep semantic understanding through contextual embeddings. BERT is pre-trained on massive amounts of unlabeled text data using tasks like “Masked Language Modeling” (predicting masked words) and “Next Sentence Prediction,” allowing it to learn deep contextualized representations of words and sentences, and fine-tunes on specific tasks, excelling in tasks such as question answering, named entity recognition, and document classification.⁵⁰ Unlike conventional NLP methods that rely on rule-based systems or shallow statistical models, BERT uses transformer-based deep neural networks to capture the full context of a word based on its surroundings—both left and right. This contrasts with earlier models like Word2Vec or GloVe, which generate static embeddings and treat words the same regardless of context.^50–52

BERT, while excellent for understanding word context, is not directly optimized for sentence similarity. SBERT (Sentence-BERT) is a modification of the pre-trained BERT model (and similar Transformer models), specifically designed for generating sentence embeddings that accurately reflect semantic similarity, making it more suitable for clustering tasks. SBERT achieves this by fine-tuning BERT using a siamese or triplet network architecture, enabling it to produce embeddings where the distance between them corresponds to semantic similarity, making it a more direct and efficient approach.^53,54 The introduction of models like BERT and SBERT has massively improved document clustering and semantic analysis, leading to significantly better quality and handling capabilities. These advancements allow us to understand and organize large text datasets with unprecedented accuracy and efficiency.^53,55–61

With the approaches and scope defined above, the study is guided by the following Research Objectives (ROs).

RO1: To assess research productivity in the intersection domain of sustainability and technology within food systems, validating trends and scholarly engagement.

RO2: To systematically map the research landscape, identifying the scope, depth, and dominant themes in sustainability and technological advancements within food systems.

RO3: To analyze key contributors, influential research networks, and their impact on shaping discourse and innovation in the field.

RO4: To uncover emerging trends and latent themes by applying deep document clustering techniques for advanced literature analysis.

In alignment with these objectives, the study addresses the following Research Questions (RQs).

RQ1: What is the current state of research productivity at the intersection of sustainability, technology, and food system? This question examines academic output, publication trends, and the evolution of research contributions in the field.

RQ2: How comprehensively have sustainability and technology been explored within food systems, and what are the dominant research themes? This question seeks to map the breadth and depth of scholarly work across sustainability dimensions (e.g., environmental, economic, social) and technological advancements (e.g., AI, optimization).

RQ3: Who are the leading contributors in this domain, and how do their research networks influence the field? This question investigates prominent authors, institutions, and research collaborations, identifying their role in shaping the discourse and practical applications of sustainability and technology in food systems.

RQ4: What emerging trends and underlying themes can be identified through deep document clustering techniques? This question employs advanced computational methods to detect latent patterns, offering insights into potential future research directions and interdisciplinary integration.

To answer these questions, the paper is structured as follows: Section 2 explains the methodology and design of the research. The results are then presented in Section 3. Section 4 is a discussion of the results. Section 5 concludes with the main findings, final remarks, future research, and limitations.

This study provides a major contribution to capture the landscape of today’s food systems, particularly on how sustainability challenges intersect with rapid technological growth, through multi-tiered literature search strategy.^2,3 Unlike traditional literature reviews, the strategy enables us to quantify publications which are normally masked by traditional single objective approach. For instance, prior studies have focused on individual objectives such as high level conceptual or managerial perspectives, for example green supply chain management framework^62,63 or LCA on environmental sustainability of food processing¹¹ or even social-political context,^64,65 majority with particular application of new technology to the system, such as IOT, deep learning and blockchain in SCM and planning,^24,66,67 application of machine learning to LCA⁶⁸ and agriculture technology adoption,^69,70 this research takes a more balanced, yet unique, interdisciplinary perspective with boarder keyword range covers popular and key concept well representing the context, to uncover patterns across commercial, technological, and sustainability-driven research. By answering our research questions, we create a complete, multi-faceted sketch of this segment.

By applying deep document clustering to bibliometric analysis and comparing these findings with those from conventional analytical approaches reveals that how data is processed can significantly change the story it tells, showing both the capabilities and drawbacks of different methods.

Overall, our work improves upon existing knowledge by thoroughly examining the less-studied conditions within the sustainability, technology, and food systems nexus, thereby connecting previously separate research efforts.

Methodology

Design scope for the study

This study focuses on three contexts of food system: the commercial (or the economic) contexts, the sustainability contexts, and the technological contexts within the food system. A multi-tiered literature search strategy^2,3 is employed, focusing on specific key terms within each area. The terms are brainstormed and filtered internally with team members, and further cross check and verify with the number of publications with mere the terms from Scopus (data not shown), where the terms and selection are designed to match the criterias of: 1. Terms selected have to be general enough, yet specific, to represent or could represent the relevance of the document content to the context Definition 2. Selected terms collectively can capture the maximum space of the defined context 3. Terms are general, and layman enough to be highly possible chosen by general writer to represent the concept of the context.

• The Commercial Context: This encompasses the demand and supply of food to the food system, terms shortlisted are “food supply,” “food product,” “food process,” “food system*”, and “food management.” This scope encompasses the entire food value chain, from upstream farming to midstream manufacturing and downstream distribution.

• The Sustainability Context: This focuses on sustainability concepts, including terms such as “sustain*”, “ESG,” “life cycle assessment,” “management,” “eco*”, and “circular.”

• The Technological Context: This focuses on the rapidly arising information technology in AI, ML, and optimization techniques. Terms shortlisted are “deep learning,” “unconstrained optimization,” “constrained optimization,” “natural language processing,” “artificial intelligence,” “large language model,” “objective optimization,” and “machine learning.”

The multi-tiered literature search strategy comprehensively explore the interplay between these critical contexts within the food systems (Table 1).

Table 1.

The context, keywords, definitions and references of the keywords used.

Keywords	Definitions	References
Commercial context
“Food supply”	The food supply encompasses the entire network of processes and activities involved in producing, processing, distributing, and making food available to consumers. This includes agricultural production, post-harvest handling, processing, packaging, storage, transportation, wholesale and retail distribution, and consumption	^71,72
“Food product*”	Food products are any substances or items intended for human consumption, whether processed, partially processed, or unprocessed. They encompass various goods, including raw agricultural commodities, processed foods, beverages, and prepared meals	⁷³
“Food process*”	Food processing encompasses the set of methods and techniques used to transform raw agricultural materials into consumable food products. These processes can include physical, chemical, and biological transformations, such as cleaning, sorting, milling, heating, cooling, fermentation, packaging, and preservation	^73,74
“Food system*”	A food system encompasses all the elements (environment, people, inputs, processes, infrastructure, institutions, etc.) and activities that relate to the production, processing, distribution, preparation, and consumption of food, as well as the outputs of these activities, including socio-economic and environmental outcomes	¹
“Food management”	Food management refers to the planning, organization, direction, and control of activities along food production and supply chain, from farm to consumer. This includes managing resources, processes, logistics, quality, safety, and sustainability aspects to ensure effective and safe food production and distribution	^75–77
Sustainability context
“Sustain*”	Representing sustainability concept, refers to the ability to meet present needs without compromising the ability of future generations to meet their own needs, integrating environmental, social, and economic dimensions	⁸
“ESG”	ESG stands for environmental, social, and governance. It comprises a set of criteria used to evaluate a company’s ethical impact and sustainability practices, which in turn influence investment decisions and long-term performance	⁷⁸
“Life cycle assessment”	Life cycle assessment (LCA) is a systematic method for evaluating the environmental impacts of a product, process, or service throughout its entire life cycle—from raw material extraction to end-of-life disposal	⁷⁹
“Management”	Representing keyword to capture management activity in the food value chain including the process of planning, organizing, leading, and controlling an organization’s resources to achieve its objectives efficiently and effectively	^80,81
“Eco*”	The prefix “eco*” is used as a wildcard to capture terms related to ecology and environmental considerations—such as eco-friendly, ecological, or eco-efficient—emphasizing aspects of environmental sustainability	NA
“Circular”	Representing keyword capturing circular economy concept, which is an economic system that aims to eliminate waste and promote the continual use of resources through practices such as recycling, reuse, and sustainable product design	⁸²
Technological context
“Deep learning”	A subfield of machine learning that uses multi-layered neural networks to automatically learn hierarchical representations from data, achieving state-of-the-art performance on complex tasks	⁸³
“Unconstrained optimi*”	The process of finding the optimum (minimum or maximum) of an objective function without any restrictions on the decision variables	⁸⁴
“Constrained optimi*”	The process of optimizing an objective function subject to restrictions (e.g., equality or inequality constraints) on the decision variables	⁸⁴
“Natural language processing”	A field of artificial intelligence focused on enabling computers to understand, interpret, and generate human language using computational and linguistic techniques	³⁶
“Large language model”	An advanced AI system, typically based on deep learning architectures (e.g., transformers), trained on vast amounts of text data to understand and generate human-like language across various contexts	⁸⁵
“Objective optimise*”	The systematic process of improving an objective function—by either maximizing or minimizing its value—through iterative adjustments of decision variables, usually within a mathematical programming framework	⁸⁴
“Machine learning”	A branch of artificial intelligence that focuses on developing algorithms that enable computers to learn from data and progressively improve performance without being explicitly programmed for each task	⁸⁶
“Artificial intelligence”	The broad field of computer science devoted to creating systems capable of performing tasks that typically require human intelligence, including learning, reasoning, and problem-solving	Russell, S. J., & Norvig, P. (2021)⁸⁷

Research design and configurations for multi-tiered literature search strategy

The primary database for this review is Scopus, chosen for its extensive coverage of peer-reviewed literature across multiple disciplines. To ensure comprehensive retrieval of relevant studies, Boolean operators (AND, OR) are used to construct multiple search strings that integrate commercial, sustainability, and technological contexts generating all relevant keyword combinations following the Cartesian product. The search is conducted within the title, abstract, and keywords fields, covering articles published up to 2023.

To refine results, specific exclusion criterias are applied: only finalized journal articles are considered, excluding book chapters and conference proceedings. Additionally, only articles published in English or Chinese are retained. The initial search is expected to yield 2,254 articles, which are further filtered based on the exclusion criteria, reducing the dataset to 988 articles (Table 2, Figure 1).

Table 2.

Keywords for Context, Boolean Configurations for Combining Search string and search results.

Commercial context	Sustainability context	Technological context
“Food supply”	“Sustain*”	“Deep learning”
“Food product*”	“ESG”	“Unconstrained optimi*”
“Food process*”	“Life cycle assessment”	“Constrained optimi*”
“Food system*”	“Management”	“Natural language processing”
“Food management”	“Eco*”	“Artificial intelli*”
	“Circular”	“Large language model”
		“Objective optimise*”
		“Machine learning”
Boolean Configurations for Combining Search string for		search results
(Commercial context term AND sustainability context term AND technological context term) OR (commercial context term AND sustainability context term AND (another technological context term)) OR …		2254 articles

Figure 1.

Research Design and Scheme of analysis.

The output is carried forward for bibliomeric analysis using Bibliometrix, through Biblioshiny, a web-interface app for Bibliometrix.³³ and procedure for deep language document clustering as per described by Makan et al.²⁷

Findings

Publications productivity

The bibliometric analysis of publication trends from 1987 to 2023 reveals a significant growth in research output over time (Figure 2). Initially, the number of articles published was extremely low, with only a few sporadic publications between 1987 and 2007. A notable increase began in 2008, when 11 articles were published, followed by moderate but consistent growth in subsequent years. The research output experienced a sharp rise starting in 2017 (21 articles), followed by a rapid acceleration in 2020 (79 articles). The most significant growth occurred between 2021 and 2023, with the number of publications rising from 148 in 2021 to 276 in 2023, marking the peak of research activity in the analyzed period.

Figure 2.

Annual distribution of publications from 1987 to 2023.

Most Influential Sources

The bibliometric analysis of the most influential sources highlights the key journals contributing to the research landscape (Table 3). Remote Sensing is the most influential journal, with 48 published articles, followed by Computers and Electronics in Agriculture (30 articles) and Journal of Cleaner Production (23 articles). Notably, Elsevier and MDPI dominate the publishing landscape, with Elsevier contributing extensively through high-impact journals such as Science of the Total Environment (23 articles), International Journal of Production Economics (7 articles), and Journal of Environmental Management (8 articles). Switzerland and the Netherlands emerge as leading publishing hubs, with MDPI and Elsevier collectively producing the majority of influential publications.

Table 3.

Most influential sources.

Rank	Journal title	Subject area	Publishing Company	Country	Scopus cite score	Scopus CiteScore highest quartile	Number of articles
1	Remote sensing	Remote sensing	MDPI	Switzerland	6	Q1	48
2	Computers and electronics in agriculture	Agriculture, electronics	Elsevier	Netherlands	6.8	Q1	30
3	Journal of cleaner production	Environmental science, engineering	Elsevier	Netherlands	11.1	Q1	23
4	Science of the total environment	Environmental science	Elsevier	Netherlands	10.2	Q1	23
5	Sustainability (Switzerland)	Environmental science, social sciences	MDPI	Switzerland	3.9	Q2	19
6	Agricultural water management	Agriculture, water resources	Elsevier	Netherlands	6.7	Q1	15
7	IEEE access	Multidisciplinary	IEEE	USA	4.5	Q1	15
8	Sensors	Engineering, instruments	MDPI	Switzerland	3.9	Q2	15
9	Nongye Gongcheng Xuebao/Transactions of the Chinese society of agricultural engineering	Agricultural engineering	CSAE	China	1.8	Q4	14
10	Agriculture (Switzerland)	Agriculture	MDPI	Switzerland	3.4	Q2	11
11	ISPRS journal of photogrammetry and remote sensing	Remote sensing, photogrammetry	Elsevier	Netherlands	10.5	Q1	10
12	Expert systems with applications	Artificial intelligence, computer science	Elsevier	Netherlands	8.7	Q1	9
13	Applied sciences (Switzerland)	Multidisciplinary	MDPI	Switzerland	3.2	Q3	8
14	Energies	Energy	MDPI	Switzerland	4.7	Q1	8
15	Journal of environmental management	Environmental science, management	Elsevier	Netherlands	8.6	Q1	8
16	Scientific reports	Multidisciplinary	Nature publishing group	UK	4.4	Q1	8
17	Agronomy	Agriculture	MDPI	Switzerland	3.5	Q2	7
18	Environmental research letters	Environmental science	IOP publishing	UK	6.5	Q1	7
19	Foods	Food science	MDPI	Switzerland	4.7	Q2	7
20	International journal of production economics	Production, economics	Elsevier	Netherlands	8.9	Q1	7

Publications by country

The bibliometric analysis of publications by country (Table 4) reveals that China leads in research output with 558 publications, followed by the USA (400) and India (243). European countries such as the UK (115), Italy,⁸⁹ and the Netherlands⁴⁹ also contribute significantly. Among Middle Eastern nations, Iran⁸² and Saudi Arabia⁴⁵ have notable research contributions. Malaysia and Mexico both have 37 publications, while South Korea,³³ Egypt,²⁸ and Greece²⁸ reflect emerging research efforts in their respective regions.

Table 4.

Publications by country.

Region	Publications
China	558
USA	400
India	243
UK	115
Italy	89
Iran	82
Australia	64
France	56
Germany	56
Netherlands	49
Spain	47
Canada	45
Saudi Arabia	45
Brazil	42
Pakistan	39
Malaysia	37
Mexico	37
South Korea	33
Egypt	28
Greece	28

Rank	Author(s)	Year	DOI	Study title	Total citations	Publication journal
1	Wang G	2017	10.1155/2017/2917536	Automatic image-based plant disease severity estimation using deep learning	519	Computational intelligence and Neuroscience
2	Govindan K	2014	10.1016/j.ijpe.2013.12.028	Two-echelon multiple-vehicle location–routing problem with time windows for optimization of sustainable supply chain network of perishable food	503	International journal of production economics
3	Sharma R	2020	10.1016/j.cor.2020.104926	A systematic literature review on machine learning applications for sustainable agriculture supply chain performance	371	Computers & operations research
4	Klerkx L	2020	10.1016/j.gfs.2019.100347	Dealing with the game-changing technologies of agriculture 4.0: How do we manage diversity and responsibility in food system transition pathways?	297	Global food security
5	Wei CL	2010	10.1371/journal.pone.0015323	Global patterns and predictions of seafloor biomass using random forests	281	PLOS ONE
6	Saurabh S	2021	10.1016/j.jclepro.2020.124731	Blockchain technology adoption, architecture, and sustainable agri-food supply chains	265	Journal of cleaner production
7	Validi S	2014	10.1016/j.ijpe.2014.02.003	A case analysis of a sustainable food supply chain distribution system—A multi-objective approach	225	International journal of production economics
8	Zhang C	2008	NA	Automatic keyword extraction from documents using conditional random fields	211	Journal of computer information systems
9	Lee SH	2020	10.1016/j.compag.2020.105220	New perspectives on plant disease characterization based on deep learning	201	Computers and electronics in agriculture
10	Bottero M	2011	10.1016/j.envsoft.2011.04.002	Application of the analytic hierarchy process and the analytic network process for the assessment of different wastewater treatment systems	192	Environmental modelling & software

Most prominent topics

The bibliometric analysis of top keywords by frequency of occurrence reveals most prominent topics in sustainability, technological advancements, management systems, and commercial applications in the agricultural and food supply chain sectors (Figure 3, Table 6).

Figure 3.

Word cloud from the analysis.

Table 6.

Top keywords by frequency of occurrence.

Relevant context	Keyword(s)	Occurrences
Sustainability	Food supply	404
Technological	Machine learning	284
Technological	Deep learning	187
Sustainability	Crops	184
Technological	Artificial intelligence	166
Sustainability	Agriculture	160
Sustainability	Food security	146
Technological	Remote sensing	136
Sustainability	Sustainable development	121
Management system	Learning systems	118
Commercial	Forecasting	116
Management system	Decision making	111
Sustainability	Climate change	110
Others	China	94
Others	Article	86
Technological	Multiobjective optimization	83
Management system	Decision support systems	81
Sustainability	Water supply	78
Technological	Learning algorithms	72
Technological	Optimization	72
Commercial	Food production	70
Technological	Machine-learning	70
Sustainability	Cultivation	69
Technological	Decision trees	66
Commercial	Crop yield	65
Technological	Agricultural robots	63
Sustainability	Land use	63
Sustainability	Water management	63
Technological	Algorithm	62
Sustainability	Human	58
Sustainability	Irrigation	56
Commercial	Crop production	54
Sustainability	Forestry	52
Technological	Random forests	50
Technological	Support vector machines	50
Sustainability	Economic and social effects	47
Commercial	Supply chains	46
Technological	Mapping	45
Sustainability	Vegetation	45
Technological	Classification (of information)	44
Sustainability	Humans	43
Technological	Neural networks	43
Technological	Regression analysis	43
Sustainability	Environmental impact	42
Commercial	Supply chain management	42
Technological	Satellite imagery	41
Technological	Convolutional neural networks	40
Sustainability	Economics	40
Commercial	Food processing	40
Sustainability	Farms	38

Sustainability Context: “Food supply” (404 occurrences) is the most frequently mentioned keyword, reflecting the significance of ensuring sustainable and efficient food distribution. Other high-frequency sustainability terms include “crops” (184), “agriculture” (160), “food security” (146), “sustainable development” (121), “climate change” (110), and “water supply”.⁷⁸ These indicate a strong research focus on balancing agricultural productivity with environmental and social concerns. Topics such as “land use”,⁶³ “water management”,⁶³ “forestry”,⁵² and “environmental impact”⁴² highlight broader ecological considerations.

Technological Context: “Machine learning” (284) and “deep learning” (187) are prominent keywords, showing the increasing role of AI-driven analytics in food and agriculture research. Other related terms such as “artificial intelligence” (166), “remote sensing” (136), “learning algorithms”,⁷² “decision trees”,⁶⁶ and “neural networks”⁴³ emphasize the growing reliance on computational techniques for predictive analytics and automation. “Optimization”,⁷² “multi-objective optimization”,⁸³ and “algorithm”⁶² highlight the mathematical and engineering methodologies applied to agricultural and food production systems.

Commercial Context: “Food production”,⁷⁰ “crop yield”,⁶⁵ and “crop production”⁵⁴ emphasize research directed at improving agricultural efficiency and maximizing output. “Supply chains”,⁴⁶ “supply chain management”,⁴² and “food processing”⁴⁰ illustrate the critical role of logistics and production in ensuring food sustainability.

It is noteworthy that the frequent appearance of “decision making” (111) and “decision support systems”⁸¹ suggests that many studies focus on improving efficiency through structured analytical frameworks. “Learning systems” (118) and “forecasting” (116) are also recurrent terms, indicating a research interest in predictive modeling and adaptive management strategies. The findings highlight the prominent role of the management system and decision support system.

Other notable keywords including “China” (94 occurrences) appear frequently, suggesting a significant research contribution from Chinese institutions. Terms like “economic and social effects”,⁴⁷ “economics”,⁴⁰ and “human”⁵⁸ underscore the socio-economic dimensions of agricultural sustainability.

Thematic Map Network: The co‐occurrence analysis reveals two major thematic clusters that capture the multifaceted research landscape at the intersection of food supply, sustainability, and technology (Figure 4, Figure 5).

Figure 4.

Thematic clusters from thematic analysis.

Figure 5.

Thematic Network from thematic analysis.

Cluster 1: Food Supply This cluster is dominated by high-frequency keywords related to agriculture and technological applications within food systems, serving as the basic themes in the field (Figure 4, Figure 6). The term “food supply” appears 387 times and holds a high PageRank centrality (0.04,398), indicating its pivotal role in the network. Other prominent keywords include “machine learning” (194 occurrences) and “deep learning” (129), which underscore the integration of advanced AI techniques into agricultural practices. Additional terms such as “crops” (182), “remote sensing” (98), “learning systems” (118), and “forecasting” (103) further highlight the focus on precision agriculture, predictive analytics, and resource management. Notably, high betweenness centrality scores for keywords such as “deep learning” (51.92) and “remote sensing” (47.90) reveal that these technologies serve as crucial bridges that connect various subfields within this cluster. Other terms like “cultivation,” “decision trees,” and “crop yield” emphasize operational aspects and process optimization, which are essential for enhancing food security and sustainability.

Cluster 2: Artificial Intelligence This cluster centers on the broader application of AI in addressing sustainability challenges, acting as a niche themes by the timeframe the study is carried out. High-frequency keywords include “artificial intelligence” (135 occurrences), accompanied by terms such as “agriculture” (114), “food security” (119), “sustainable development” (114), and “decision making” (98). These terms suggest that AI is increasingly applied to optimize resource allocation, support decision-making, and drive sustainable practices across food systems. The betweenness centrality values, particularly for “decision making” (114.48) and “sustainable development” (84.88), indicate that these concepts are key connectors linking various interdisciplinary topics. Moreover, methodological terms such as “multiobjective optimization” (77) and “decision support systems” (81) further emphasize the analytical approaches used to address complex, multi-criteria sustainability challenges.

Collaboration Network

The bibliometric analysis reveals key authors contributing to the research domain, classified into multiple clusters (Figure 6). Cluster 1 has the most dominant presence, with authors such as Wang X, Wang Y, and Li Y showing high betweenness and PageRank values, indicating their strong influence and connectivity in the research network. Cluster 2 and Cluster 3 contain additional notable contributors like Zhang W, Zhang X, and Sun Y, demonstrating moderate influence within their respective groups. Other clusters, such as Clusters 4, 5, and 6, highlight a more dispersed network of researchers, while Cluster 7 exhibits high PageRank scores, suggesting a focused but impactful presence. Clusters 9 and 10 include authors with unique characteristics, such as Azzaro-Pantel C and Aguilar-Lasserre AA, showing closeness values of 1, indicating their specialized but potentially impactful research contributions.

Figure 6.

Collaboration Network.

Clusters analyzed from deep document clustering technique

The document clustering analysis, conducted across the domains of commercial, sustainability, and technology, yielded six distinct thematic clusters, each characterized by unique underlying themes and focus areas (Figure 7, Table 7). The first cluster, designated as Label 0 (Global Systems and Demand Challenges), adopts a macroeconomic perspective that emphasizes global and regional demand, system models, and systemic challenges, addressing large-scale economic and logistical issues such as demand forecasting and resource distribution. In contrast, Label 1 (Methodological and Problem-Solving Approaches) is highly technical, focusing on problem-solving, optimization, and the application of fuzzy logic, and is distinguished by its emphasis on analytical tools and decision-making methodologies critical for navigating complex, uncertain environments in both commercial and technological contexts. The second cluster, Label 2 (Environmental Practices and Supply Chain Sustainability), centers on sustainability within supply chains by highlighting environmental practices, carbon footprint management, and supplier relationships, thereby reflecting the growing importance of integrating sustainability into procurement and supply chain operations. Similarly, Label 3 (Sustainable Logistics and Distribution) addresses operational challenges, particularly in logistics and supply chain solutions, with an emphasis on sustainable transportation and distribution systems. Label 4 (Precision Agriculture and Data-Driven Approaches) shifts the focus toward agriculture, emphasizing precision farming, data analytics, and sustainable land use, which are essential for improving agricultural productivity and addressing food security challenges. Finally, Label 5 (Energy and Industrial Technology) explores industrial applications and energy systems, with a focus on energy efficiency, technological innovation, and decision-making in industrial settings, particularly emphasizing renewable energy and sustainable industrial practices.

Figure 7.

Clustering result from document clustering.

Table 7.

Clusters analyzed from deep document clustering technique.

Most frequent words	Label 0 (Global systems and demand challenges)	Label 1 (Methodological and problem-solving approaches)	Label 2 (Environmental practices and supply chain sustainability)	Label 3 (sustainable logistics and distribution)	Label 4 (precision agriculture and data-driven approaches)	Label 5 (Energy and industrial technology)
1	Demand	Method	Environmental	Distribution	Crop	Energy
2	System	Objective	Practice	Supply	Approach	Process
3	Model	Problem	Cf	Chain	Accuracy	Technology
4	Global	Fuzzy	Supplier	Solution	Using	Decision
5	Regional	Function	Selection	Sustainable	Information	Support
6	Challenge	Mating	Contract	Performance	Assessment	Selecting
7	Increase	Solve	Order	Approach	Land	Thermal
8	Cropping	Total	Gain	Presented	System	Food
9	Motivated	New	Performance	Transportation	Proposed	Industry
10	Food	Presented	Step	Decision	Predict	Sustainability

Discussion

Below findings provide sufficient information in answering RQ1

Scientific Publications: The observed trend in publication output indicates a delayed but exponential growth in research activity within the field. The near absence of publications before 2008 suggests that either the topic is not a major research focus or the technological and methodological advancements required to support significant academic contributions have not yet been developed. The increase in publications from 2008 onward likely reflects growing interest in the field, driven by technological advancements, increased awareness of sustainability issues, or policy shifts encouraging research in related domains.

A major turning point occurred after 2017, where publication numbers grew sharply, indicating a phase of accelerated research and knowledge dissemination. The 21st Conference of the Parties in Paris in 2015, Parties to the UNFCCC reached a landmark agreement to combat climate change and to accelerate and intensify the actions and investments needed for a sustainable low carbon future marked a significant progress, follow by the exponential rise from 2020 onward suggests that recent global challenges, such as climate change, digital transformation, outbreak of Covid-19 and the following supply chain disruptions, may have stimulated research efforts. The surge in 2021-2023 could also be attributed to the increasing adoption of advanced methodologies, such as artificial intelligence, data analytics, and automation, especially years around 2023 when OpenAI in AI began to soar in popularity. Overall, the result has confirmed the hypothesis for RQ1. However, it is worth noting that the number of publications at the peak of 2023 at 276 publications are still considered low, compared with hot topics which can easily exceed 20k publications annually, from Scopus database only (data not shown). This indicates that the segment remains underexplored by the moment this study is carried out.

Most Influential Sources: The dominance of Remote Sensing as the most influential journal suggests a strong research focus on geospatial technologies, remote sensing applications, and environmental monitoring. The presence of Computers and Electronics in Agriculture in the second position reflects the increasing role of digitalization, artificial intelligence, and automation in modern agricultural practices.

The prevalence of Elsevier and MDPI as key publishers indicates the continue gorwing importance of open-access and high-impact publishing platforms in the field. Switzerland, as the home of MDPI, and the Netherlands, as the base of Elsevier, reinforce the European influence in scholarly research and publication hub, in contrast to Publications by Country, where researches are led by China, the USA and India.

Journals such as Journal of Cleaner Production and Science of the Total Environment emphasize sustainability and environmental science, demonstrating a shift towards integrating green technologies and sustainable practices across industries. The inclusion of multidisciplinary journals such as IEEE Access and Scientific Reports highlights the interdisciplinary nature of current research, bridging engineering, environmental science, and computational approaches.

Overall, the findings indicate a well-balanced distribution of research across remote sensing, environmental science, agriculture, and digital transformation, reflecting the growing integration of technology in addressing sustainability challenges, which align with the hypothesis.

Publications by Country: The dominance of China in research publications aligns with its increasing investment in technological advancements and scientific research. The USA, maintaining its position as a global research leader, follows closely, demonstrating a strong research ecosystem driven by top-tier universities and institutions. India’s ranking in third place highlights its growing presence in academic research, particularly in fields such as agriculture, environmental science, and sustainable development.

European nations, including the UK, Italy, and Germany, maintain a substantial contribution, reflecting their strong research infrastructure and international collaborations. Iran’s presence in the top 10 suggests significant advancements in scientific research, despite geopolitical and economic challenges.

The contributions from Malaysia, Mexico, and South Korea indicate emerging research efforts from developing economies, demonstrating a shift towards global knowledge production. Meanwhile, Middle Eastern countries such as Saudi Arabia and Iran reflect regional efforts to strengthen academic and scientific capabilities.

Overall, we found that other than the traditional research house such as China and the USA continue to lead, emerging economies are increasingly contributing to the global scientific landscape. The rsults also indicate that, while Western countries remain as the leading hub for publications, China and India are leading many emerging countries catching up on this.

Below findings provides sufficient information in answering RQ2

Top Cited Original Articles: The analysis of the most cited articles provides valuable insights into research trends and the growing influence of artificial intelligence, sustainability, and supply chain optimization in modern scientific studies.

The top two most cited papers highlight the intersection of AI with agriculture and supply chain management. The success of Wang (2017) and Sharma (2020) underscores the increasing reliance on deep learning and machine learning for real-time decision-making and efficiency improvements in food production and logistics. AI-driven applications in plant disease detection (Wang, 2017; Lee, 2020) are particularly impactful in advancing precision agriculture and minimizing losses.

Govindan (2014) and Validi (2014) demonstrate the role of advanced logistics models in sustainable food distribution, particularly for perishable goods. This aligns with the rising global emphasis on reducing food waste and enhancing efficiency within supply chains. Blockchain technology (Saurabh, 2021) further adds to this discourse, offering decentralized solutions for improved traceability and transparency in agri-food systems.

The discussion by Klerkx (2020) on Agriculture 4.0 highlights the increasing complexity of managing diverse emerging technologies in food systems. His work stresses the need for strategic governance in transitioning towards more sustainable and responsible agricultural practices.

The contributions by Wei (2010) and Bottero (2011) indicate the importance of machine learning in ecological predictions and environmental decision-making. With increasing global concerns over resource management, these studies reinforce the role of AI-driven analytics in sustainability efforts.

The work by Zhang (2008) demonstrates early advancements in NLP for automatic keyword extraction, which is foundational in modern AI-driven data processing and information retrieval systems.

The highly cited articles collectively reflect the evolution of AI, supply chain optimization, and sustainability research. The increasing citation counts of studies in these areas indicate their growing relevance in tackling real-world challenges related to agriculture, food supply chains, and environmental management. Moving forward, integrating emerging technologies such as AI, blockchain, and decision-making frameworks will be crucial in shaping the future of sustainable food production and logistics. The result also reflect the diversity of the research interest within the research areas, seeing a focus on agriculture study, implying that other research areas remain underexplored with big rooms of opporutnity for improvements.

Word Cloud: The keyword analysis provides valuable insights into the dominant research areas, showcasing the interplay between sustainability, technological innovations, decision-making frameworks, and commercial applications.

Sustainability as a core theme

The prominence of sustainability-related keywords highlights the ongoing global emphasis on food security, resource conservation, and climate change adaptation. The frequent occurrence of “food supply” (404) and “food security” (146) suggests that researchers are prioritizing ways to ensure stable and sustainable food distribution. Additionally, concerns over “water management”⁶³ and “land use”⁶³ indicate the necessity of balancing agricultural expansion with environmental preservation.

The rise of AI and automation in agriculture

The strong presence of machine learning (284), deep learning (187), and artificial intelligence (166) demonstrates the rapid adoption of computational techniques in agricultural research. AI-driven methodologies are being leveraged for tasks such as remote sensing (136), predictive modeling (learning algorithms: 72), and automated decision-making (decision trees: 66). The application of optimization techniques⁷² further emphasizes the push for efficiency in food production and supply chain logistics.

The importance of decision support systems

The presence of keywords like “decision making” (111) and “decision support systems”⁸¹ suggests that many studies focus on enhancing strategic planning and operational efficiency. Forecasting (116) is also a key topic, indicating that predictive models play a crucial role in managing uncertainties in agriculture and food supply chains.

Economic and commercial considerations

The frequent mention of “food production”,⁷⁰ “crop yield”,⁶⁵ and “supply chains”⁴⁶ suggests a strong commercial orientation in agricultural research. The focus on supply chain management⁴² highlights the importance of logistics and efficiency in ensuring food availability. Additionally, keywords like “economic and social effects”⁴⁷ and “economics”⁴⁰ reflect broader considerations of financial sustainability and social impact.

Regional and contextual relevance

The high occurrence of “China”⁹⁴ suggests that a substantial portion of research originates from or focuses on China, reinforcing the country’s role as a leading contributor to agricultural and technological studies. The mention of “human”⁵⁸ and “humans”⁴³ suggests that studies are also exploring the human dimensions of sustainability, particularly in terms of labor, policy-making, and socio-economic impacts.

The bibliometric keyword analysis indicates that sustainability, technological advancements, and efficient decision-making frameworks are the dominant themes in agricultural and food supply chain research. The increasing role of AI and machine learning suggests a paradigm shift towards automation and predictive analytics, while the focus on supply chain management underscores the importance of logistics and efficiency in food production. Moving forward, research in this area is likely to continue integrating advanced technologies with sustainability strategies to address global food security challenges.

Co-occurrence Analysis: On the other hand, co‐occurrence analysis provides a robust understanding of how research in food systems, sustainability, and technology is structured into distinct yet interconnected thematic areas.

In Cluster 1, the prominence of terms like “food supply,” “machine learning,” and “deep learning” illustrates the critical role of technology in modern agriculture. These keywords reveal that recent research has increasingly focused on integrating AI techniques into agricultural operations to improve crop management, optimize resource use, and enhance overall food security. The high betweenness centrality of deep learning and remote sensing underscores their role as bridges, facilitating the flow of information between traditional agricultural practices and innovative technological methods. This cluster clearly reflects an operational shift, where traditional production metrics are now complemented by advanced, data-driven approaches to meet the evolving challenges of food security in a sustainable manner.

In Cluster 2, the focus shifts towards the strategic and systemic application of AI to sustainability challenges. Here, keywords such as “artificial intelligence,” “sustainable development,” and “decision making” illustrate how AI is not merely a technical tool but a strategic enabler that helps address complex, multidisciplinary issues. The integration of multiobjective optimization and decision support systems indicates that researchers are developing sophisticated frameworks to balance economic, environmental, and social objectives. This cluster highlights the transition from conventional production-centric approaches to more holistic, integrated strategies that are essential for sustainable development. The high connectivity of these terms suggests that AI-driven methodologies are crucial for designing systems that can navigate trade-offs and uncertainties inherent in sustainability challenges.

Interestingly, commercial aspects like “forecasting,” “food processing,” and “supply chain management” suggest a focus on optimizing food production and distribution. The presence of “multiobjective optimization” and “economic and social effects” indicates efforts to balance sustainability with economic viability, ensuring that food supply strategies remain cost-effective while minimizing environmental impact.

Overall, the findings from the co‐occurrence network analysis underscore the dual role of technology in food systems. On one hand, technological innovations—particularly in AI, machine learning, and remote sensing—are driving operational improvements in agriculture, as evidenced in Cluster 1. On the other hand, these technologies are being leveraged at a strategic level to address broader sustainability challenges, as shown in Cluster 2. This multifaceted approach reflects the evolving nature of research in this area, where interdisciplinary integration is key. The results not only validate the growing influence of AI in optimizing food production and resource management but also emphasize the need for continued collaboration across disciplines to further refine and implement these technologies for sustainable development.

These insights have significant implications for both academic research and practical applications. For researchers, the analysis provides a roadmap for exploring underdeveloped areas, such as advanced AI methodologies for sustainability optimization. For practitioners and policymakers, the findings highlight the potential of integrating data-driven approaches to improve decision-making and operational efficiency within food systems. Implying more comprehensive models that can better predict, manage, and optimize the complex interactions between technology and sustainability in the global food industry for future research.

While sections make a good sketch for RQ3

Collaboration Network Analysis: The results suggest that certain authors, particularly in Cluster 1, play a central role in shaping the research field, likely acting as key knowledge hubs. Their high betweenness values indicate their importance in connecting various research groups. Meanwhile, authors in Clusters 2 and 3 exhibit moderate influence, possibly contributing to subdomains within the broader research field. Clusters 4 to 7 show a more distributed network, with some authors having higher PageRank scores, signifying their citation importance despite lower betweenness values. This suggests a mix of emerging researchers and established authors with niche expertise. The presence of authors in Clusters 9 and 10 with unique closeness characteristics implies that their research may be highly specialized, potentially contributing significant insights within specific subfields rather than acting as intermediaries.

These findings indicate a well-structured research landscape with both dominant contributors and emerging researchers playing essential roles in advancing the field.

For RQ4 which aims to further uncover emerging trends and latent themes by applying deep document clustering techniques for advanced literature analysis. The clustering results reveal significance in scope, focus, and domain emphasis, underscoring the multidisciplinary nature of the intersection among commercial, sustainability, and technology domains. The macroeconomic orientation of Label 0 contrasts sharply with the technical and methodological focus of Label 1, while Labels 2 and 3 reflect a convergence of environmental and operational concerns in supply chain sustainability and logistics, respectively. Labels 4 and 5 further illustrate the importance of industry-specific applications, with Label 4 emphasizing the role of precision agriculture and data-driven approaches in enhancing food production and security, and Label 5 highlighting the critical function of energy-efficient and innovative technologies in industrial settings. These findings underscore the necessity for tailored, interdisciplinary approaches that integrate sustainability into commercial and technological practices in order the capture the entire landscape. They also suggest that future research should address the unique challenges and opportunities within each thematic area, thereby fostering innovation and supporting strategic and operational improvements in organizations operating within a contemporary global business environment. The increasing complexity of global food supply chains, climate change, and resource management necessitates an integrated, multidisciplinary approach that bridges engineering, business, and management disciplines. This study, based on a thematic map analysis, identifies two major clusters—Food Supply (Label 1) and Artificial Intelligence (Label 2)—both of which highlight the transformative potential of AI-driven technologies in enhancing sustainability, optimizing operations, and driving strategic advancements in the global business environment. The findings underscore the critical role of AI in decision-making, operational efficiency, and policy integration, aligning with the journal’s objective of promoting holistic, sustainable technology management in business.

Distinct clustering result from BERT

The application of deep document clustering has yielded a markedly different and more distinct thematic structure compared to traditional bibliometric methods such as those implemented through Biblioshiny. While Biblioshiny primarily relies on top keyword frequencies and co-word networks (as seen in Table 6 and Figures 4 and 5), these methods often result in overlapping clusters with more ambiguous thematic boundaries. For example, take the top 3 most frequent keywords from the two clusters using Biblioshiny, Cluster 1 = food supply, machine learning and crops, Cluster 2 = artificial intelligence, food security and agriculture, they are highly similar and overlapped which does not differentiate the theme from each other. In contrast, deep NLP-based clustering techniques extracted six thematically rich and well-separated clusters, each characterized by unique and context-sensitive keyword sets. This indicates a higher granularity and interpretive clarity than the two broad and intersecting clusters derived from Biblioshiny.

This discrepancy highlights the power of NLP in semantic text analysis, which does not merely count keywords but interprets their contextual relationships, co-dependencies, and syntactic structures. NLP-based clustering, particularly when leveraging transformer models such as BERT or SBERT, captures subtle semantic distinctions that traditional term-frequency methods often overlook. Similar effects are observed in other studies, with NLP showcasing its strong semantic analytical potential within and beyond bibliometric analysis field^{38,39,42,48,55,88,89}

Given the growing scale and complexity of scholarly corpora, traditional bibliometric tools may no longer suffice in capturing the full thematic richness and interdisciplinary evolution. NLP-based document clustering is poised to become not just a complementary approach, but a core analytical method in future bibliometric analyses. Its ability to semantically segment large textual datasets into non-overlapping, coherent themes provides significant advantages in mapping intellectual structures, detecting emerging research areas, and supporting strategic decision-making for funding and policy development.

How emerging technologies and sustainability practices shape the food systems?

The outcomes of the deep document clustering and bibliometric analysis strongly align with the central objective of this study—to explore the landscape of thematic evolution and interconnections between sustainability, technological advancement, and commercial food systems. The resulting clusters reveal an increased concentration of research at the intersections of these high-impact domains, indicating that this space is gaining traction. However, the relatively low volume of output within some intersectional clusters suggests that while interest is growing, the research landscape remains underexplored, leaving significant room for scholarly contribution. This is particularly relevant in interdisciplinary zones where sustainable development and digital transformation converge with operational and policy dimensions of food systems.

In addition, despite the global relevance of food systems, the bibliometric network reveals that a large portion of scholarly outputs and collaborations are concentrated in three countries: China, the United States, and India. These nations dominate the field in terms of publication volume and author centrality, indicating that they have strategically aligned national research agendas and policy frameworks to reinforce research and innovation in agri-tech, food security, and sustainable development. For example, China’s 14th Five-Year Plan, the U.S. Farm Bill 2023, and India’s National Mission on Sustainable Agriculture have placed innovation at the forefront of food system transformation. This geographical consolidation, while productive, also suggests potential risks related to regional research bias and a lack of representation from Global South or low-income countries, where food system vulnerabilities are most pronounced.⁹⁰

Furthermore, although the document clustering analysis revealed a diverse range of research interests, a significant portion of publications focus primarily on agriculture—particularly topics such as crop performance, agroecology, and environmental input efficiency. While agriculture is indeed foundational, this narrow focus may overshadow other critical components of the food system, including post-harvest processing, cold chain logistics, food retail innovation, operations strategies, and food waste management. It is well recognized that sustainability challenges are increasingly emerging not just at the farm level, but across the entire food value chain, from input sourcing to consumer behavior.⁹¹ This underscores a need for broader and more balanced research that explores system-wide sustainability transformations and techno-commercial innovations across diverse operational contexts.

In sum, the current research landscape reflects both progress and fragmentation: while advances in NLP-based bibliometric techniques are uncovering clearer thematic boundaries and untapped intersections, the field still suffers from thematic and geographical concentration. There is a critical need for more integrative, system-wide, and globally representative research that explores how commercial imperatives and sustainability goals can be aligned across the entire food system—not just in agricultural production but across logistics, policy, retail, and consumption.

Conclusions

This study advances our understanding of how sustainability imperatives and rapid technological developments are reshaping the commercial food systems landscape. By employing advanced bibliometric techniques—especially natural language processing (NLP), deep document clustering, and BERT-based models—this research reveals clearer thematic distinctions and more granular knowledge structure than traditional methods like Biblioshiny alone. The clustering analysis confirmed that while research interest in the intersection of sustainability, technology, and food systems is growing, significant thematic and geographic gaps remain.

The dominance of contributions from countries like China, the United States, and India suggests strategic policy-driven investment in sustainable food system innovations. However, the thematic concentration on agriculture—while important—indicates a pressing need for further exploration into broader areas of the food value chain, including logistics, retail, consumer behavior, and circular practices. Moreover, the integration of sustainability in commercial settings remains a complex challenge, especially where economic feasibility and policy alignment are still maturing.

Overall, this research demonstrates the critical potential of NLP-driven bibliometric analysis to uncover hidden patterns, emerging themes, and knowledge gaps. It provides both theoretical implications for future academic inquiry and practical insights for policy makers and industry stakeholders seeking to navigate the complex interface of sustainability and technology in food systems. As sustainability becomes not just a regulatory expectation but a competitive differentiator, aligning research, policy, and industry strategy will be pivotal in building resilient, responsible, and future-ready food systems.

Limitation and future work

Despite offering valuable insights, this study has several limitations that should be acknowledged. First, the bibliometric analysis is based solely on the Scopus database, which, while comprehensive, may have excluded relevant publications indexed in Web of Science, PubMed, IEEE Xplore, or Google Scholar. This single-source reliance may introduce publication bias and limit the generalizability of the findings across broader disciplines and geographies.

Second, the search string is developed through internal brainstorming rather than systematic keyword harvesting or expert recommendation procedure. As a result, potentially relevant documents may have been omitted due to variations in terminology, especially in an interdisciplinary field that spans sustainability, food systems, and technological innovation.

Third, the temporal scope and citation-based metrics used in the clustering methods may underrepresented emerging research, particularly recently published work that has not yet accumulated citations.

Furthermore, this study primarily focused on co-authorship, co-citation, and keyword co-occurrence patterns. Future research could benefit from integrating altmetric indicators, patent data, or policy documents to offer a more holistic view of influence and practical application. Also, integrating geospatial analysis and institutional affiliations could shed light on how knowledge flows across global innovation systems.

Future studies are encouraged to adopt multi-database strategies, use a more systematic keyword expansion techniques, and apply other clustering models.^92–94 Such advancements could yield richer, more accurate representations of knowledge landscapes, particularly in fast-evolving and multi-sectoral domains like food systems.

Footnotes

ORCID iDs

Chee Perng Ng

Tiong Yew Tang

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

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

References

Nguyen

. Sustainable food systems: concept and framework. FAO, 2018.

Bramer

de Jonge

Rethlefsen

, et al. A systematic approach to searching: an efficient and complete method to develop literature searches. J Med Libr Assoc 2018; 106(4): 531–541.

Gusenbauer

Gauster

. How to search for literature in systematic reviews and meta-analyses: a comprehensive step-by-step guide. Technol Forecast Soc Change 2025; 212: 123833.

Evenson

Gollin

. Assessing the impact of the green revolution, 1960 to 2000. Science 2003; 300(5620): 758–762.

Clark

Domingo

NGG

Colgan

, et al. Global food system emissions could preclude achieving the 1.5° and 2°C climate change targets. Science 2020; 370(6517): 705–708.

Tilman

Balzer

Hill

, et al. Global food demand and the sustainable intensification of agriculture. Proc Natl Acad Sci 2011; 108(50): 20260–20264.

WWF . Building a nature-positive society. CBS Library, 2022, p. 1.

World Commission on Environment and Development . Our common future. Oxford University Press, 1987, p. 383.

Lisboa

Pasquali

dos Anjos

, et al. Innovative and sustainable food preservation techniques: enhancing food quality, safety, and environmental sustainability. Sustainability 2024; 16(18): 8223. https://www.mdpi.com/2071-1050/16/18/8223

10.

Notarnicola

Tassielli

Renzulli

, et al. Environmental impacts of food consumption in Europe. J Clean Prod 2017; 140: 753–765.

11.

Chung

MMS

Bao

Zhang

, et al. Life cycle assessment on environmental sustainability of food processing. Annu Rev Food Sci Technol 2022; 13: 217–237.

12.

Cahyadi

Hidayati

Zahra

, et al. Integrating circular economy principles into agri-food supply chain management: a systematic literature review. Sustainability 2024; 16(16): 7165.

13.

De Bernardi

Bertello

Forliano

. Circularity of food systems: a review and research agenda. Br Food J 2023; 125(3): 1094–1129.

14.

Hamam

Chinnici

Di Vita

, et al. Circular economy models in agro-food systems: a review. Sustainability 2021; 13(6): 3453. https://www.mdpi.com/2071-1050/13/6/3453

15.

Munonye

. Towards circular economy metrics: a systematic review. In Circular economy and sustainability. Springer Nature 2025.

16.

Vann Yaroson

Chowdhury

Mangla

, et al. A systematic literature review exploring and linking circular economy and sustainable development goals in the past three decades (1991–2022). Int J Prod Res 2024; 62(4): 1399–1433.

17.

European Commission . Farm to fork strategy 2020. https://food.ec.europa.eu/system/files/2020-05/f2f_action-plan_2020_strategy-info_en.pdf

18.

Usda.gov . U.S. food loss and waste 2030 champions 2024. https://www.usda.gov/about-food/food-safety/food-loss-and-waste/us-food-loss-and-waste-2030-champions

19.

Rahman

deb

Alam

, et al. Navigating the contemporary landscape of food waste management in developing countries: a comprehensive overview and prospective analysis. Heliyon 2024; 10(12): e33218.

20.

Shen

Hong

, et al. The status of the global food waste mitigation policies: experience and inspiration for China. Environ Dev Sustain 2023: 1–29.

21.

Aggarwal

Srinivas

Puppala

, et al. Integrated decision support for promoting crop rotation based sustainable agricultural management using geoinformatics and stochastic optimization. Comput Electron Agric 2022; 200: 107213.

22.

Banasik

Kanellopoulos

Claassen

GDH

, et al. Assessing alternative production options for eco-efficient food supply chains using multi-objective optimization. Ann Oper Res 2017; 250(2): 341–362.

23.

Deb

. Multiobjective optimization using evolutionary algorithms. Wiley, 2001.

24.

Hosseinnia Shavaki

Ebrahimi Ghahnavieh

. Applications of deep learning into supply chain management: a systematic literature review and a framework for future research. Artif Intell Rev 2023; 56(5): 4447–4489.

25.

Kamilaris

Fonts

Prenafeta-Boldύ

. The rise of blockchain technology in agriculture and food supply chains. Trends Food Sci Technol 2019; 91: 640–652.

26.

Kamilaris

Prenafeta Boldú

. Deep learning in agriculture: a survey. Comput Electron Agric 2018; 147: 16.

27.

Rakmakan. GitHub . GitHub - rakmakan/clustering-with-bert: powerful document clustering models are essential as they can efficiently process large sets of documents. These models can be helpful in many fields, including general research. Searching through large corpora of publications can be a slow and tedious task; such models can significantly reduce this time 2021. https://github.com/rakmakan/Clustering-with-BERT

28.

Pritchard

. Statistical bibliography or bibliometrics? J Doc 1969; 25: 348–349.

29.

Broadus

. Toward a definition of bibliometrics. Scientometrics 1987; 12(5): 373–379.

30.

Donthu

Kumar

Mukherjee

, et al. How to conduct a bibliometric analysis: an overview and guidelines. J Bus Res 2021; 133: 285–296.

31.

Mukherjee

Lim

Kumar

, et al. Guidelines for advancing theory and practice through bibliometric research. J Bus Res 2022; 148: 101–115.

32.

Öztürk

Kocaman

Kanbach

. How to design bibliometric research: an overview and a framework proposal. Rev Manag Sci 2024; 18: 3333–3361.

33.

Aria

Cuccurullo

. Bibliometrix: an R-tool for comprehensive science mapping analysis. J Informetr 2017; 11(4): 959–975.

34.

Singapore Management University (SMU) . Bibliometrix – a powerful and popular new bibliometric tool used in the domain of business and management. Singapore Management University (SMU), 2022.

35.

Cobo

López-Herrera

Herrera-Viedma

, et al. Science mapping software tools: review, analysis, and cooperative study among tools. J Am Soc Inf Sci Technol 2011; 62(7): 1382–1402.

36.

Jurafsky

Martin

. Speech and language processing: an introduction to natural language processing. Computational linguistics, and speech recognition. Prentice Hall, 2009, p. 988.

37.

Manning

Schütze

. Foundations of statistical natural language processing. MIT Press, 1999, p. 680.

38.

Chen

Tsang

. When text mining meets science mapping in the bibliometric analysis: a review and future opportunities. Int J Eng Bus Manag 2023; 15: 18479790231222349.

39.

Chen

Xie

Wang

, et al. A bibliometric analysis of natural language processing in medical research. BMC Med Inf Decis Making 2018; 18(S1): 14.

40.

Chen

Tian

Fang

. Bibliometric analysis of natural language processing using CiteSpace and VOSviewer. Nat Lang Process J 2025; 10: 100123.

41.

Cobelli

Blasi

. Combining topic modeling and bibliometric analysis to understand the evolution of technological innovation adoption in the healthcare industry. Eur J Innovat Manag 2024; 27(9): 127–149.

42.

Dsouza

Leyens

Brand

. Employing bibliometrics and natural language processing (NLP) to analyse real-world applications of adverse drug reaction. Explor Res Clin Soc Pharm 2025; 18: 100592.

43.

Meng

Zhang

Huang

, et al. Topic discovery via latent space clustering of pretrained language model representations. In Proceedings of the ACM Web Conference 2022. Association for Computing Machinery, 2022, pp. 3143–3152.

44.

Ogunleye

Lancho Barrantes

Zakariyyah

. Topic modelling through the bibliometrics lens and its technique. Artif Intell Rev 2025; 58(3): 74.

45.

Paladugu

Kancherla

. Harnessing deep learning techniques for text clustering and document categorization. Int J Recent Innov Trends Comput Commun 2023; 11(8): 125–139.

46.

Pessin

Santos

CAS

Yamane

, et al. A method of mapping process for scientific production using the smart bibliometrics. MethodsX 2023; 11: 102367.

47.

Petukhova

Matos-Carvalho

Fachada

. Text clustering with large language model embeddings. Int J Cogn Comput Eng 2025; 6: 100–108.

48.

Sandu

Cotfas

Stănescu

, et al. A bibliometric analysis of text mining: exploring the use of natural language processing in social media research. Appl Sci (Basel) 2024; 14(8): 3144. https://www.mdpi.com/2076-3417/14/8/3144

49.

Sun

Meng

, et al. Sentence similarity based on contexts. Trans Assoc Comput Linguist 2022; 10: 573–588.

50.

Devlin

Chang

Lee

, et al. BERT: pre-training of deep bidirectional transformers for language understanding. arXiv 2019. https://arxiv.org/abs/1810.04805

51.

Grootendorst

. BERTopic: neural topic modeling with a class-based TF-IDF procedure 2022. https://arxiv.org/abs/2203.05794

52.

Sun

Qiu

, et al.

How to fine-tune BERT for text classification?

2020. https://arxiv.org/abs/1905.05583

53.

Ortakci

. Revolutionary text clustering: investigating transfer learning capacity of SBERT models through pooling techniques. Eng Sci Technol Int J 2024; 55: 101730.

54.

Reimers

Gurevych

. Sentence-BERT: sentence embeddings using siamese BERT-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). Association for Computational Linguistics, 2019, pp. 3980–3990. https://www.aclweb.org/anthology/D19-1410

55.

George

Sumathy

. An integrated clustering and BERT framework for improved topic modeling. Int J Inf Technol 2023; 15(4): 2187–2195.

56.

Mehta

Agarwal

Kaliyar

. A comprehensive and analytical review of text clustering techniques. Int J Data Sci Anal 2024; 18(3): 239–258.

57.

Mersha

yigezu

Kalita

. Semantic-driven topic modeling using transformer-based embeddings and clustering algorithms. arXiv 2024.

58.

Ortakci

Borhan

. Optimizing SBERT for long text clustering: two novel approaches with empirical insights. J Supercomput 2025; 81(8): 950.

59.

Sampaio

Maxcici

. Unsupervised document and template clustering using multimodal embeddings. arXiv 2025. https://arxiv.org/abs/2506.12116

60.

Tshitoyan

Dagdelen

Weston

, et al. Unsupervised word embeddings capture latent knowledge from materials science literature. Nature 2019; 571(7763): 95–98.

61.

Tussupov

Kassymova

Mukhanova

, et al. Analysis of short texts using intelligent clustering methods. Algorithms 2025; 18(5): 289. https://www.mdpi.com/1999-4893/18/5/289

62.

Dubey

Gunasekaran

Papadopoulos

, et al. Sustainable supply chain management: framework and further research directions. J Clean Prod 2017; 142: 1119–1130.

63.

Mavani

Ali

Othman

, et al. Application of artificial intelligence in food Industry—A guideline. Food Eng Rev 2022; 14(1): 134–175.

64.

Béné

Abdulai

. Navigating the politics and processes of food systems transformation: guidance from a holistic framework. Front Sustain Food Syst 2024. https://www.frontiersin.org/journals/sustainable-food-systems/articles

65.

Brons

van der Gaast

Awuh

, et al. A tale of two labs: rethinking urban living labs for advancing citizen engagement in food system transformations. Cities 2022; 123: 103552.

66.

Tan

, et al. Dynamic blockchain adoption for freshness-keeping in the fresh agricultural product supply chain. Expert Syst Appl 2023; 217: 119494.

67.

Tsang

Lam

, et al. Integrating internet of things and multi-temperature delivery planning for perishable food E-commerce logistics: a model and application. Int J Prod Res 2021; 59(5): 1534–1556.

68.

Ghoroghi

Rezgui

Petri

, et al. Advances in application of machine learning to life cycle assessment: a literature review. Int J Life Cycle Assess 2022; 27(3): 433–456.

69.

Klerkx

Rose

. Dealing with the game-changing technologies of agriculture 4.0: how do we manage diversity and responsibility in food system transition pathways? Global Food Secur 2020; 24: 100347.

70.

Kotte

, et al. Artificial intelligence (AI) and its applications in agriculture: a review. Environ Conserv J 2024; 25(1): 274–288.

71.

FAO . Urbanization, agrifood systems transformation and healthy diets across the rural-urban continuum. FAO, 2023, p. 283. The state of food security and nutrition in the world.

72.

The state of food security and nutrition in the world 2022. FAO; 2022. https://www.fao.org/documents/card/en/c/cc0639en

73.

Fellows

. Food processing technology: principles and practice. 5th ed., Woodhead Publishing, 2022, p. 779.

74.

Singh

Heldman

. Introduction to food engineering. 5th ed., Elsevier, Academic Press, 2014. Food science and technology international series.

75.

Bolton

. Quality management systems for the food industry: a guide to ISO 9001/2; [companion volume to practical approaches to food control and food quality series]. Blackie Acad. & Professional, 1997, vol 1, p. 193.

76.

Dani

. Food supply chain management and logistics: understanding the challenges of production, operation and sustainability in the food industry. 2nd ed. : Kogan Page Limited, 2021, p. 1.

77.

Food control management | food safety and quality. Food and Agriculture Organization of the United Nations. https://www.fao.org/food-safety/food-control-systems/official-controls/food-control-management/en/

78.

Eccles

Ioannou

Serafeim

. The impact of corporate sustainability on organizational processes and performance. Manag Sci 2014; 60(11): 2835–2857.

79.

Finnveden

Hauschild

Ekvall

, et al. Recent developments in life cycle assessment. J Environ Manag 2009; 91(1): 1–21.

80.

Koontz

O’Donnell

Weihrich

. Essentials of management. 4th ed., McGraw-Hill, 1986, p. 564. (McGraw-Hill series in management).

81.

Pullman

. Food supply chain management: building a sustainable future. 2nd ed., Routledge, 2021.

82.

Potocnik

. Towards a circular economy: business rationale for an accelerated transition. 2015.

83.

Goodfellow

Bengio

Courville

. Deep learning, The MIT Press, 2016, p. 775. (Adaptive computation and machine learning).

84.

Nocedal

Wright

. Numerical optimization. 2nd ed., Springer, 2006, p. 1. (Springer series in operations research and financial engineering).

85.

Brown

Mann

Ryder

, et al.

Language models are few-shot learners

2020. https://arxiv.org/abs/2005.14165

86.

Janiesch

Zschech

Heinrich

. Machine learning and deep learning. Electron mark. 2021. Septentrion; 31(3): 685–695.

87.

Russell

Norvig

. Artificial intelligence: a modern approach. Pearson, 2016, p. 1.

88.

Mersha

yigezu

Kalita

. Semantic-driven topic modeling using transformer-based embeddings and clustering algorithms. arXiv 2024.

89.

Tanev

Sieklicki

. Using topic modeling as a semantic technology: examining research article claims to identify the role of non-human actants in the pursuit of scientific inventions. Appl Sci 2025; 15(6): 3253.

90.

The state of food and agriculture. FAO; 2021. https://www.fao.org/documents/card/en/c/cb4476en

91.

Fanzo

Covic

Dobermann

, et al. A research vision for food systems in the 2020s: defying the status quo. Global Food Secur 2020; 26: 100397.

92.

Google.com. 2025. https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=all+minilm+l6+v2&oq=all-MiniLM-L6+c

93.

Huggingface.co . Models - hugging face 2025. https://huggingface.co/models?pipeline_tag=sentence-similarity&sort=trending

94.

huggingface.co . Using sentence transformers at hugging face. https://huggingface.co/docs/hub/en/sentence-transformers

How urgent environmental challenges and tech advancements reshape commercial food systems: A bibliometric analysis leveraging natural-language processing

Abstract

Keywords

Introduction

The surge of call for sustainability and ESG

The rise of new information processing technology

Impact to food production

Purpose of the study

Bibliometric analysis

Current tools: Bibliometrix and biblioshiny

Natural language processing (NLP) and its incorporation into bibliometric analysis

BERT and SBERT: Transforming text mining in bibliometrics

Methodology

Design scope for the study

Research design and configurations for multi-tiered literature search strategy

Findings

Publications productivity

Most Influential Sources

Publications by country

Top cited original articles

Most prominent topics

Collaboration Network

Clusters analyzed from deep document clustering technique

Discussion

Below findings provide sufficient information in answering RQ1

Below findings provides sufficient information in answering RQ2

Sustainability as a core theme

The rise of AI and automation in agriculture

The importance of decision support systems

Economic and commercial considerations

Regional and contextual relevance

While sections make a good sketch for RQ3

Distinct clustering result from BERT

How emerging technologies and sustainability practices shape the food systems?

Conclusions

Limitation and future work

Footnotes

ORCID iDs

Funding

Declaration of conflicting interests

References