My research contributions are multifaceted and encompass successful industry project outcomes, concurrent management of multiple projects, collaborations at national and international levels, supervision of higher degree research students, and production of high-quality publications. These achievements serve as a testament to the impact of my research, both nationally and internationally.
In particular, my successful industry project outcomes reflect the value of my contributions to practical applications. Concurrent management of multiple projects highlights my ability to effectively deliver complex initiatives. Collaborations on a national and international level have facilitated knowledge-sharing and relationships that have advanced research in various fields.
As a result of my diverse research contributions, my work has had national and international impacts, underscoring the significance of my research outcomes. Notable examples of my successful projects are as follows.
17. Predictive Modeling for Bioinformatics:
[2023-present] As a QUT research fellow, I am collaborating with Associate Prof Leila Cuttle, Principal Research Fellow and head of the QUT Burns and Trauma Laboratory. This project focuses on enhancing biomedical prediction using multi-modal omics-centric data. By integrating information from diverse sources such as DNA, mRNA, miRNA, and patient demographics, the study aims to improve burn healing prediction accuracy. We proposed a Gaussian Process-based learning framework to address challenges related to modality bias and ensures that more informative modalities contribute significantly to the overall representation. This research was awarded with QUT CDS Second Byte Fund 2024.
16. Digitising Scanned Cadastre Survey into Structured Data:
[2023-2024] As a QUT research fellow, I worked with Bennett + Bennett, a company specializing in surveying and related services. The project aimed to automate the digitisation of scanned land cadastral surveys using intelligent algorithms. It focused on converting scanned cadastral plans into vector data and storing this information in a structured JSON format. Key features extracted included OCRed text, lines, tables, reference marks, plan descriptions, lot parcels, and zoom areas.
A proof-of-concept (POC) software was developed, demonstrating the system's ability to accurately identify and extract these features, and store them in a structured format. The POC was tested on 121 survey plans, successfully meeting the project's objectives.
15. Industry Report Value Extraction:
2022-2023] As a QUT research fellow, I am worked with the Queensland Department of Resources' Geological Survey of Queensland (GSQ). GSQ has amassed a large collection of industry reports as a result of exploration and production companies' legal obligations in the state. This project created automated tools and techniques utilizing language models based on deep learning, to extract essential information from these reports in the form of question answer. The extracted information will aid in future resource exploration endeavours. I have published the following paper under this project.
[1] Talukder, A. S., Nayak, R., & Bashar, M. A. (2023, October). GAN-IE: Generative Adversarial Network for Information Extraction with Limited Annotated Data. In International Conference on Web Information Systems Engineering (pp. 633-642). Singapore: Springer Nature Singapore.
14. Multi Modal Data Representation:
[2021-present] As QUT research fellow I am working on the project of integrating multi-modal data in deep learning models that can offer valuable and diverse insights into complex scenarios. However, combining data from multiple modalities to form a feature representation poses a challenge due to bias in features within or between modalities. Current multi-modal fusion methods struggle to learn generic features because of varying noise patterns and feature dynamics across modalities. This study aims to investigate techniques for effectively integrating features from different modalities to create a unified representation for downstream tasks, such as classification. I have published the following papers under this project. This research was awarded with QUT CDS Second Byte Fund 2023.
[1] Luong, K., Nayak, R., Balasubramaniam, T., & Bashar, M. A. (2022). Multi-layer manifold learning for deep non-negative matrix factorization-based multi-view clustering. Pattern Recognition, 131, 108815.
[2] Zhang, D., Wang, Y., Bashar, M. A., & Nayak, R. (2023, May). Enhanced Topic Modeling with Multi-modal Representation Learning. In Pacific-Asia Conference on Knowledge Discovery and Data Mining (pp. 393-404).
[3] Zhang, D., Nayak, R., & Bashar, M. A. (2021, December). Exploring Fusion Strategies in Deep Learning Models for Multi-Modal Classification. In Data Mining: 19th Australasian Conference on Data Mining, AusDM 2021, Brisbane, QLD, Australia, December 14-15, 2021, Proceedings (pp. 102-117). Singapore: Springer Singapore.
13. Social Media for Cyber Threat Intelligence (CTI):
[2021-Now] The proactive cyber defence and mitigation of potential cyber attacks of organisations can be enhanced by collecting information about cyber threats from various sources. Twitter has been demonstrated to offer timely Cyber Threat Intelligence (CTI) regarding zero-day software vulnerabilities and exploits. Nevertheless, extracting and examining valuable insights, patterns, and trends from abundant unstructured tweets is challenging. This project aims to develop a data-driven framework that leverages unsupervised topic modelling techniques to collect, analyze, and monitor tweets. I have published the following papers under this project. This research was awarded with QUT CDS Second Byte Fund 2024.
[1] Duoyi Zhang, Yue Wang, Md Abul Bashar, Richi Nayak (2023). Enhanced Topic Modeling with Multi-modal Representation Learning. The 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD) (accepted).
[1] Wang, Y., Bashar, M. A., Chandramohan, M., & Nayak, R. (2023). Exploring topic models to discern cyber threats on Twitter: A case study on Log4Shell. Intelligent Systems with Applications, 20, 200280.
12. Repatriation of Indigenous Ancestral Human Remains:
[2021-Now] I am working with Australian National University (ANU) as a research fellow, where I am contributing to Data Science stream of a project under ARC Discovery grant in Digital Humanities. My primary objective is to create machine learning theories and models to comprehend data and information concerning human remains trading. To achieve this, I am developing algorithms that enable historians to collect, analyse, and visualize relevant data from historical events, thus assisting the Repatriation of Indigenous Ancestral Human Remains. The National Indigenous Times and QUT have published two news articles on this research. A summary of media coverage on this project can be found here. I have published the following papers under this project.
[1] Bashar, M. A., Nayak, R., Knapman, G., Turnbull, P., & Fforde, C. (2023). An Informed Neural Network for Discovering Historical Documentation Assisting the Repatriation of Indigenous Ancestral Human Remains. Social Science Computer Review, 08944393231158788.
11. Predictive Maintenance:
[2021-Now] As a research fellow at QUT, I worked with Verton, a load-management system developing company, where I contributed to a project under Innovative Manufacturing CRC grant to accelerate the commercialisation of the world’s first and ground-breaking technology to manage suspended loads. I developed models for the data stream to understand the usage patterns of the suspended load management system and identify anomalies for predictive maintenance. I have published the following papers under this project.
[1] Bashar, M. A., & Nayak, R. (2020, December). TAnoGAN: Time series anomaly detection with generative adversarial networks. In 2020 IEEE Symposium Series on Computational Intelligence (SSCI) (pp. 1778-1785). IEEE.
10. Updating Population Data Shift:
[2021-present] As a research fellow at QUT, I am collaborating with Australian Bureau of Statistics (ABS), where I led a research team that developed a method to update the underlying distribution of past data in ABS repository using recently conducted survey data.
9. Monitoring COVID-19 Social Media Conversation Dynamics:
[2020] As a QUT research associate, I conducted this project in partnership with QUT's Digital Observatory, where I created models capable of examining the dynamics of COVID-19 conversations on Twitter in the space of both time and location. Specifically, I developed deep learning based topic and sentiment analysis models to capture these dynamics. I have published two papers [] under this project. I have published the following papers under this project.
[1] Bashar, M. A., Nayak, R., & Balasubramaniam, T. (2022). Deep learning-based topic and sentiment analysis: COVID19 information seeking on social media. Social Network Analysis and Mining, 12(1), 90.
[2] Balasubramaniam, T., Nayak, R., Luong, K., & Bashar, M. A. (2021). Identifying Covid-19 misinformation tweets and learning their spatio-temporal topic dynamics using Nonnegative Coupled Matrix Tensor Factorization. Social Network Analysis and Mining, 11(1), 57.
[3] Balasubramaniam, T., Nayak, R., & Bashar, M. A. (2020, December). Understanding the spatio-temporal topic dynamics of covid-19 using nonnegative tensor factorization: a case study. In 2020 IEEE symposium series on computational intelligence (SSCI) (pp. 1218-1225). IEEE.
[4] Shahriar, F., & Bashar, M. A. (2021). Automatic Monitoring Social Dynamics During Big Incidences: A Case Study of COVID-19 in Bangladesh. arXiv preprint arXiv:2101.09667.
8. Bias Detection:
[2020] As a research associate at QUT, I worked with iShield.ai, a deep-tech US company, where I created a comprehensive bias detection method using deep learning. This product was later commercialized by iShield.ai and is now being used to help Fortune 500 companies manage their content. The product can be integrated with Slack, a messaging app for business, and is currently available on Slack Store. I have published the following papers under this project.
[1] Bashar, M. A., Nayak, R., Kothare, A., Sharma, V., & Kandadai, K. (2021). Deep Learning for Bias Detection: From Inception to Deployment. In Data Mining: 19th Australasian Conference on Data Mining, AusDM 2021, Brisbane, QLD, Australia, December 14-15, 2021, Proceedings 19 (pp. 86-101). Springer Singapore.
7. Early Classification of Longitudinal data:
[2020-present] As a QUT research fellow, I am working in a project, namely Classifying longitudinal data early on (ELC). ELC is difficult due to missing data caused by irregular sampling and drop-outs. To address this issue, I am leading a research team that created a new model called the Longitudinal Early Classifier GAN (LEC-GAN), which is context-conditional and inspired by the generative capabilities of the Generative Adversarial Network (GAN). The model incorporates informative missingness, static features, and class embeddings to enhance the ELC objective. I have published the following papers under this project.
[1] Pingi, S. T., Bashar, M. A., & Nayak, R. (2022, December). A Comparative Look at the Resilience of Discriminative and Generative Classifiers to Missing Data in Longitudinal Datasets. In Data Mining: 20th Australasian Conference, AusDM 2022, Western Sydney, Australia, December 12–15, 2022, Proceedings (pp. 133-147). Singapore: Springer Nature Singapore.
[2] Pingi, S. T., Zhang, D., Bashar, M. A., & Nayak, R. (2023). Joint Representation Learning with Generative Adversarial Imputation Network for Improved Classification of Longitudinal Data. Data Science and Engineering, 1-21.
6. Propensity-to-Pay:
[2019] As a research associate at QUT, I collaborated with Energy Queensland (EQ) and created a model based on Bayesian Neural Network (BNN) to forecast customers' likelihood of paying their electricity bills, while also determining the degree of uncertainty in the data distribution. EQ places significant importance on estimating uncertainty because acting based on forecasts incurs a cost, while not taking any action has consequences. I have published the following papers under this project.
[1] Bashar, M. A., Nayak, R., Astin-Walmsley, K., & Heath, K. (2023). Machine Learning for Predicting Propensity-to-Pay Energy Bills. Intelligent Systems with Applications, 200176.
5. Online Abuse Detection
[2018-2022] As research a associate at QUT, I worked with QUT Institute for Future Environment (IFE) and Centre for Data Science (CDS), where I developed a transfer learning-based method for detecting misogynistic tweets with small training set. This method was featured in more than 200 national and international articles including Forbes, 7News, Brisbane Times, Daily Mail, Times of India, and Hindustan Times. A list of coverage can be found here and here. I have published the following papers under this project.
[1] Bashar, M. A., & Nayak, R. (2021). Active learning for effectively fine-tuning transfer learning to downstream task. ACM Transactions on Intelligent Systems and Technology (TIST), 12(2), 1-24.
[2] Bashar, M. A., Nayak, R., Luong, K., & Balasubramaniam, T. (2021). Progressive domain adaptation for detecting hate speech on social media with small training set and its application to COVID-19 concerned posts. Social Network Analysis and Mining, 11, 1-18.
[3] Bashar, M. A., Nayak, R., & Suzor, N. (2020). Regularising LSTM classifier by transfer learning for detecting misogynistic tweets with small training set. Knowledge and Information Systems, 62, 4029-4054.
[4] Cunningham, S., Laundon, M., Cathcart, A., Bashar, M. A., & Nayak, R. (2022). First, do no harm: automated detection of abusive comments in student evaluation of teaching surveys. Assessment & Evaluation in Higher Education, 1-13.
[5] Bashar, M. A., Nayak, R., Suzor, N., & Weir, B. (2019). Misogynistic tweet detection: Modelling cnn with small datasets. In Data Mining: 16th Australasian Conference, AusDM 2018, Bahrurst, NSW, Australia, November 28–30, 2018, Revised Selected Papers 16 (pp. 3-16). Springer Singapore.
[6] Bashar, M. A., & Nayak, R. (2020). QutNocturnal@ HASOC'19: CNN for Hate Speech and Offensive Content Identification in Hindi Language. arXiv preprint arXiv:2008.12448.
4. Robotic Marketer:
[2018-2019] As a research associate at QUT, I worked with Marketing Eye, a marketing consulting firm, where I developed a prototype of marketing strategy platform that utilizes machine learning and big data. This innovative platform, called Robotic Marketer (RM), automated the generation of marketing strategies and was delivered to clients in February 2019. Since then, RM has been adopted by various companies including SAP, Oracle, Hungry Jacks, and many more, and has proven to be a successful marketing tool. QUT has published an article on Medium featuring this project, and the article can be read here.
3. Interpreting Discovered Knowledge:
[2017] As a research associate at QUT, I was involved in a project funded by the ARC Discovery grant that developed algorithms that to uncover knowledge from texts and enhance the interpretation of the knowledge by improving semantic understanding. For example, text mining techniques often generate vast amounts of knowledge that lacks semantic information, making it difficult to interpret and utilise the knowledge. In response to this challenge, I have devised a framework that mines a personalized ontology by mapping the extracted knowledge to an ontology and relevant contextual information.
[1] Bashar, M. A., & Li, Y. (2018). Interpretation of text patterns. Data Mining and Knowledge Discovery, 32, 849-884.
[2] Gao, Y., Li, Y., Lau, R. Y., Xu, Y., & Bashar, M. A. (2017). Finding semantically valid and relevant topics by association-based topic selection model. ACM Transactions on Intelligent Systems and Technology (TIST), 9(1), 1-22.
[3] Bashar, M. A., Li, Y., Shen, Y., Gao, Y., & Huang, W. (2017). Conceptual annotation of text patterns. Computational Intelligence, 33(4), 948-979.
[4] Bashar, M. A., Li, Y., & Gao, Y. (2016, October). A framework for automatic personalised ontology learning. In 2016 IEEE/WIC/ACM International Conference on Web Intelligence (WI) (pp. 105-112). IEEE.
[5] Bashar, M. A., & Li, Y. (2017). Random set to interpret topic models in terms of ontology concepts. In AI 2017: Advances in Artificial Intelligence: 30th Australasian Joint Conference, Melbourne, VIC, Australia, August 19–20, 2017, Proceedings 30 (pp. 237-249). Springer International Publishing.
[6] Alharbi, A. S., Bashar, M. A., & Li, Y. (2018). Random-sets for dealing with uncertainties in relevance feature. In AI 2018: Advances in Artificial Intelligence: 31st Australasian Joint Conference, Wellington, New Zealand, December 11-14, 2018, Proceedings 31 (pp. 656-668). Springer International Publishing.
[7] Bashar, M. A., Li, Y., Shen, Y., & Albathan, M. (2014, August). Interpreting discovered patterns in terms of ontology concepts. In 2014 IEEE/WIC/ACM International Joint Conferences on Web Intelligence (WI) and Intelligent Agent Technologies (IAT) (Vol. 1, pp. 432-437). IEEE.
2. SMS-based Registration System for Admission Test:
[2010-2013] During my time as a lecturer at Comilla University (CoU), I worked as a key member of a team that developed an automated system that allowed students without Internet access to register for CoU admission test using text message (SMS service in cellular phone). Each year, more than 40K students used to register through this system.
1. Social networking Server, the GRID:
[2008-2010] As a key team member at Structured Data Systems Limited, I played a key role in developing the backend system for the GRID (a social networking system in South Africa). VodaCom, South Africa launched the project, and it received the "New Telecommunications Service of the Year Award-2008" in Dubai.