Rob Loke, PhD

I am currrently assistant professor data science @CMIHVA.nl and have a PhD in computer science, an MA in psychology and an MSc in computer science. My earlier research from before 2017 that is not being mentioned here touched upon pattern recognition, image processing, machine learning, robotics, and adaptive control, to mention a few in arbitrary order [1]. Since my PhD was on 3D visualisation and computer graphics, I cannot wait to further extend my current and recent CMIHVA research activities listed below that are almost all in the text domain, requiring mostly natural language processing related stuff---NLP, to the image and video domains :-)

In my CMIHVA projects, I have been quite active with MSc students in measuring the omnivalent corporate reputation construct that is commonly split up into seven core components, products/services, innovation, workplace, governance, citizenship, leadership and performance, via a social listening approach based on natural language processing (NLP) in which social media, human-resource vacancy and profile data as well as online review data on the public web for various stakeholder groups of companies is being targeted and visualized to explore insights in datastream contents, opinions, and sentiments [50]. Company stakeholder groups can be customers [3,6,7,9,11], internal stakeholders such as employees [4,5,13,14] as well as the general public [10,12] for instance. Some of the developed state of the art data science processing and modeling pipelines that have been developed are: (1) custom machine learning (ML) under supervised, unsupervised, and semi-supervised learning paradigms---for instance for sentiment analysis on the specific aspects assortment, product, service and delivery [3], for social media analysis  to identify brand advocates [29] and for detection of AI-generated fake reviews [23]; (2) generative adversarial learning for fake review and review quality detection [21,22]; (3) aspect-based sentiment analysis [6,4,7,9]; (4) generic semantic search [5,10,13,17]; (5) entity extraction, for instance, of skills from vacancy texts, for monitoring relevant corporate reputation subdimensions such as innovation for example that are at play [14]; (6) explicit knowledge representation with knowledge graphs and ontologies [15,16]; (7) generic visualisation [9,10,11,12,25]; (8) social media analysis [29]; (9) social network analysis. These are beneficial to feed some of the top tool industrial applications (numbered for convenience A to M below) in many sectors and markets like ecommerce that we see in companies around us in the professional field nowadays such as:

• A---(near) real-time visualisation dashboards in, for instance, retail [9] and labor [40], or, in fashion, with environmental, social, and corporate governance (ESG) topics [11] or dashboards that are concerned with, for example, sustainability topics that are relevant in many industry sectors including the energy sector [10]; 
• B---decision support system dashboards in, for example, telecom that are driven by supervised ML [12]; and
• C---time series prediction dashboards in collaboration with, for instance, a retail company in the private sector [25]. Time series prediction is not only relevant in marketing but also in fintech for the monitoring of companies' corporate reputation and operational risk management.

Besides the social listening approach [26], a visual listening approach [24] in which relevant online image data on the public web is dealt with is envisioned to be applied soon.  

Again, together with MSc students, I have been targeting other top applications in ecommerce as well:

• D---chatbots. In [15], an ontology-based chatbot to improve the interaction between the customer and the company is presented, and, in [16], an investigation into the influence of online reviews on corporate reputation with interactive knowledge graphs. Retrieval-augmented generation [34] is applied as well for instance for helping social media marketers that work in the aviation sector to check upon unattended greenwashing in posts before publishing [49].
• E---recommender systems. In the private sector, with Florensis BV, an online recommender system was built to assist the customer decision making process.

As well as applications with:

• F---blockchain systems. Application of blockchain technology to register ownership for instance is in general potentially relevant for challenges in fintech, logistics and supply chain management that businesses around us might encounter. In [18], a blockchain-based online review system to enhance corporate reputation has been developed. Non-Fungible Tokens (NFT) marketing [44] has been studied as well, for instance, in luxury fashion. 
• G---fake review spam detectors as an example of fraud detection that has much potential for protecting companies from harm in current online society has been targeted in my CMIHVA corporate reputation project from a general data science approach, i.e. defined as the intersection of computer science/IT, math and statistics and domains/business knowledge [2], in [8]. A more specific generative AI approach [20] for fake review detection to protect online products using generative adversarial learning has been applied in [21] and detection of AI-generated fake reviews for corporate reputation management in [23]. As a spinoff application of fake review detection, quality review classification using generative adversarial learning for the hospitality industry has been targeted in [22]. Nowadays, both human and bot written fake reviews need to be detected. 
• H---virtual assistants [51]. Agentic AI for business use cases but also for automated research use case for instance.
• I---interactive knowledge graph representation dashboards. For instance, for guiding knowledge discovery in datastreams to consumers [15,16].
• J---object detectors and recognizers in images. Computer vision is relevant to process retail products in images acquired by customer and employee smartphones in supermarkets for instance [30]. 
• K---virtual influencer and brand advocate detectors in social media. For instance, in telecom [29] and Dutch fitness industry [48].
• L--- toxic echo chamber detectors in social networks. For instance, in sports [46,43].
• M---digital twins [52].

Besides sophisticated tool development, phenomenon investigation with statistics is targeted as well. Occassionally, mixed method modeling including information from surveys related to human behavior besides information extracted from online available ratings and reviews in datasets is needed. For instance, mixed method research on customer decision making in specific sectors such as the hospitality sector [19]. Some other examples that study phenomenons at play but are not strictly falling in the mixed method category are in the world of labor [45,47], b corp certification [27], aviation [36], sports [28,35,39], politics [37] and communication [38]. The overall aim of the CMIHVA is to have state-of-the-art infrastructure with datasets and toolsets that lead to insights to the reputation construct for any sector or organization that is willing to assess, analyze or improve its performance. Next to corporate reputation, brand reputation is explicitly being targeted during the last few years as well.

In my current work with CMIHVA, I have also been active with companies directly in industry in the private sector. For instance, to mention one example, with KPMG Germany, the demand of helicopter spare parts with machine learning methods on logistics and warehouse management data was predicted for ADAC Heliservice.

Besides mentioned datastreams, marketplaces such as bol.com and amazon have been targeted as well. In [21], fake review scores of different companies have been measured on amazon.nl based on product reviews. Similar or other relevant project work is planned to be done on the bol.com marketplace.

Specific processing pipelines need commonly be quite focused to be either data or model driven; however, it is expected that in systems architectural terms a hybrid mixed approach will need to be developed in near future to due advantages.

While most of the tool applications are operated on desktop laptops and PCs, some should be operated in cloud infrastructure [4]. Edge computing and IOT infrastructure have not been targeted too much yet in the CMIHVA but are nevertheless foreseen to be beneficial because of foreseen benefits of federated learning in technology transfer on the cloud edge continuum that might boost applications that circumvent data privacy and safety issues in marketing for retailers for instance [30]. Federated learning technologically links to relevant smartphone and robotics applications that are foreseen by the CMIHVA in the actual society and business field around us. Therefore, to align with the current emphasis in industry and society on edge computing, a Raspberry Pi cluster was recently purchased by CMIHVA. The cluster in development is inclusive a full operational Beowulf to be able to parallelize heavy-duty algorithms. This is relevant to set up projects that are related to edge computing applications that are booming nowadays around us, for instance in future supermarkets. The cluster also will facilitate automatic scraping of custom online web scraper tools and dashboarding of big datastreams such as job vacancies for instance.

Via the academic collaboration and established partnership with Brightdata as well as the Bright Initiative that I have established within the CMIHVA as one of the first universities worldwide in 2019, we want to stimulate original research proposals and fruitful ideas that should get explored for doing good with data [32]. 

Projects

The corporate reputation (CR) project that is currrently entitled Exploring online reputation that is led by Rob has been a successful project in the last 7 years in which multidisciplinary teams ranging from 1, 4, 6, 8 to 13 members have been working on a diverse plethora of research subjects. Since research can focus on different stakeholder groups, such as customers, employees etc., and on different sectors and markets, such as e-commerce, retail, telecom, fashion, energy, banking, aviation, sports, politics etc., as well as address different topics, e.g., sustainability, inclusivity etc., also, even in conflict situations for instance due to war [38], many research options can be addressed that can be prominently placed in contemporary societal actuality. In pioneering work by one master project student, 7 years ago, large datasets were scraped from well-known data platforms and reputation [31] measures were derived on four aspects that led to a scientific contribution in the academic literature [3]. After that, the project has been booming, yielding several published conference papers [4,5,6,7,8,10,11,12,13,15,23], conference research abstracts [9] and prospective journal papers in coauthorship with students. 

The Exploring contingent changes in labor project that explores contemporary changes in labor that is led by Rob has been another successful project in the last couple of years with prospective scientific papers coauthored with students [14,40,45,47]. 

The Exploring toxicity in social media project that is led by Rob as well. For example, toxicity around influencers was herein recently targeted [41,42].

I am also running a Computer vision prototyping for supermarket virtual assistance project in which computer vision is prototyped and applied for leveraging innovative technology for enhanced consumer choice in fresh produce [30]. The goal is to enable applications in which the shopping of customers in supermarkets is assisted with smartphone enabled technology that is driven by computer vision at its core. It also targets computer vision in a federated learning setting [33].

References

[1] https://scholar.google.com/citations?user=T5oG2HEAAAAJ

[2] https://en.wikipedia.org/wiki/Data_science

[3] Sentiment polarity classification of corporate review data with a bidirectional Long-Short Term Memory (biLSTM) neural network architecture https://doi.org/10.5220/0009892303100317 (published)

[4] Assessing Corporate Reputation from Online Employee Reviews https://link.springer.com/chapter/10.1007/978-981-16-9268-0_20 (published)

[5] A Company’s Corporate Reputation through the Eyes of Employees Measured with Sentiment Analysis of Online Reviews https://www.scitepress.org/Link.aspx?doi=10.5220/0010620603780385 (published)

[6] Aspect Based Sentiment Analysis on Online Review Data to Predict Corporate Reputation https://www.scitepress.org/Link.aspx?doi=10.5220/0010607203430352 (published)

[7] Exploring Corporate Reputation based on Sentiment Polarities That Are Related to Opinions in Dutch Online Reviews https://www.scitepress.org/PublicationsDetail.aspx?ID=kHC///NTU3o=&t=1 (published)

[8] The Role of Fake Review Detection in Managing Online Corporate Reputation https://www.scitepress.org/PublicationsDetail.aspx?ID=NErJaXyEk4s=&t=1 (published)

[9] Automatic Measurement of Corporate Reputation for Retail Companies from Online Public Data on the Web (poster published with student as 1st author at conference; paper in prep)

[10] Corporate Reputation of Companies on Twitter Seen from a Sustainability Perspective https://link.springer.com/chapter/10.1007/978-981-16-9272-7_42 (published)

[11] Corporate reputation through the eyes of social responsibility in the fashion industry https://link.springer.com/chapter/10.1007/978-981-97-3698-0_41 (published)

[12] Decision Support System for Corporate Reputation based Social Media Listening using a Cross-Source Sentiment Analysis Engine https://www.scitepress.org/PublicationsDetail.aspx?ID=h2uAr7XXd5A=&t=1 (published)

[13] Presence of Corporate Reputation Cues in Company Vacancy Texts Boosts Vacancy Attractiveness as Perceived by Employees https://doi.org/10.5220/0012863400003756 (published) 

[14] Using soft skills to leverage the innovation dimension of Corporate Reputation: A human capital approach (in prep)

[15] Designing Business Related Interaction with Reviewers for Corporate Reputation Insights Using an Ontology-Based Chatbot https://link.springer.com/chapter/10.1007/978-981-96-3081-3_21 (published)

[16] An Investigation into the Influence of Online Reviews on Corporate Reputation with interactive Knowledge graphs (in prep)

[17] Semantic Search on Aspects of Corporate Reputation (in prep)

[18] The development of a blockchain-based online review system to enhance corporate reputation (in prep)

[19] Exploring the Link between Online Corporate Reputation and Consumer Decision-Making in the Hospitality Industry - A Study of Amsterdam Hotels (in prep)

[20] https://en.wikipedia.org/wiki/Generative_artificial_intelligence

[21] Fake Review Detection of Online Products using Generative Adversarial Learning (in prep)

[22] Quality review classification for the hospitality industry (in prep)

[23] Detecting AI-Generated Reviews for Corporate Reputation Management https://www.scitepress.org/PublishedPapers/2025/135720/ (published)

[24] Liu Liu, Daria Dzyabura, Natalie Mizik (2020) Visual Listening In: Extracting Brand Image Portrayed on Social Media. Marketing Science 39(4):669-686. https://doi.org/10.1287/mksc.2020.1226 

[25] Whisky Auction Price Forecasting: Tasting the Future (in prep)

[26] Westermann, A., Forthmann, J.: Social listening: a potential game changer in reputation management How big data analysis can contribute to understanding stakeholders’ views on organizations. Corp. Commun. Int. J. (2020). ISSN: 1356–3289 https://doi.org/10.1108/CCIJ-01-2020-0028 

[27] Evaluating the impact of B Corp Certification on engagement and satisfaction from the perspective of employees as internal stakeholders (in prep)

[28] The Social Scoreboard: Assessing Social Media’s Impact on Athlete and Sponsored Brands (in prep)

[29] Identifying Brand Advocates: A Neural Network Approach to Social Media Analysis (in prep)

[30] Leveraging Innovative Technology for Enhanced Consumer Choice in Fresh Produce (in prep)