My research has focused on multimodal information processing, time series modeling, multilingual and low-recources NLP, causal information extraction, lexical semantics, biomedical informatics, social media analysis, embedding-based data representation, computational linguistics and Deep learning for NLP. I am always interested in expanding my research area in other related areas. My current and past research projects are briefly described as follows:

Multimodal Information Processing

My research interest in multimodal natural language processing (NLP) has centered on addressing complex societal issues through the fusion of textual and visual information. I have contributed to the field with publications on topics such as multimodal hate speech detection, multimodal complaint detection, and multimodal fact-checking, where the integration of linguistic and visual cues enhances model performance. I've also employed advanced techniques like multitasking and contrastive learning to improve the robustness and generalization of multimodal NLP systems. Looking ahead, I aspire to delve into the exciting realms of language grounding and the development of large-scale multimodal models, aiming to further our understanding of how language and vision intersect and to advance the capabilities of multimodal AI systems for real-world applications.

  1. Abdul Aziz et al., Multitask Contrastive Learning for Multimodal Complaint Detection: A Sentiment and Emotion-Driven Approach, COLING 2025 ( Long Paper), (Submitted).
  2. Abdul Aziz et al., CSECU-DSG at CheckThat! 2023: Transformer-based Fusion Approach for Multimodal and Multigenre Check-Worthiness, CLEF 2023: Conference and Labs of the Evaluation Forum, September 18–21, 2023, Thessaloniki, Greece.
  3. Abdul Aziz et al., CSECU-DSG@ Multimodal Hate Speech Event Detection 2023: Transformer-based Multimodal Hierarchical Fusion Model For Multimodal Hate Speech Detection, Proceedings of the 6th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE), CASE@RANLP '23
  4. Abdul Aziz et al., MMTF-DES: A Fusion of Multimodal Transformer Models for Desire, Emotion, and Sentiment Analysis of Social Media Data, Elsevier Neurocomputing Journal , Vol. -, No. -, pp. -, 2025. (Accepted)

Mutilingual and Low-resource NLP

Multilingual and code-mixed task understanding is one of the most fasinating research area, where I have made some contributions through publications on topics such as taxonomic relations between multilingual arguments, multilingual complex named entity recognition, and code-mixed multitask sentiment and emotion analysis. My vision is to push the boundaries of multilingual and low-resource natural language processing by developing a unified single model capable of handling the world's diverse languages. I aspire to seamlessly integrate this work with multimodal NLP, ushering in a new era of cross-modal, multilingual understanding that extends beyond text, embracing the rich spectrum of communication in a global, interconnected world.

  1. Abdul Aziz et al., Can Contrastive Learning Enhance Code-mixed Text Understanding? A study on Multitask Sentiment and Emotion Analysis, NAACL 2025 ( Short Papers) [Submitted].
  2. Abdul Aziz et al., CSECU-DSG at SemEval-2022 Task 3: Investigating the Taxonomic Relationship Between Two Arguments using Fusion of Multilingual Transformer Models, Proceedings of the 16th International Workshops on Semantic Evaluation (SemEval '22), 2022.
  3. Abdul Aziz et al., CSECU-DSG at SemEval-2022 Task 11: Identifying the Multilingual Complex Named Entity in Text Using Stacked Embeddings and Transformer based Approach, Proceedings of the 16th International Workshops on Semantic Evaluation (SemEval '22), 2022.

Lexical Semantics

My inital research interests lies the fascinating realm of lexical semantics, where my primary emphasis resides in the field of lexical simplification. More specifically, I am deeply engaged in the study of predicting and unraveling lexical complexity, aiming to uncover ways to simplify language for improved communication and comprehension. Through my publications, I have contributed to enhancing our understanding of how language complexity can be quantified and simplified. My future ambitions in this field involve expanding my work into multilingual and low-resource natural language processing, where I have already initiated projects aimed at creating new datasets in languages beyond English. By doing so, I aspire to foster a more inclusive and comprehensive approach to lexical semantics, catering to the linguistic diversity of the global community and promoting accessibility in text comprehension across different languages.

  1. Nabila Ayman and Abdul Aziz et al., BengaliLCP: A Dataset for Lexical Complexity Prediction in the Bengali Texts, The 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING '24), 2024.
  2. Abdul Aziz et al., Leveraging Contextual Representations with BiLSTM-based Regressor for Lexical Complexity Prediction, Elsevier Natural Language Processing Journal, Vol. 05, No. 100039, 2023.
  3. Abdul Aziz et al., CSECU-DSG at SemEval-2021 Task 1: Fusion of Transformer Models for Lexical Complexity Prediction, In Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval '21), 2021.

Check-worthy Claim Prediction

Social media eases information spreading, makes information diffusion quicker, and reaches potentially more people than traditional media, in many cases regardless of the information quality. Automatic fact-checking could be a solution to warn social media users and readers or even to stop the spreading of fake news. In this project, the objective is to identify and verify the check-worthy claims. To predict the check-worthy claims, we employed vison and language transformer models. The experimental results on standard benchmark datasets show that our approach improves the performance compared to the related methods and achieved 1st position. We are currently focusing on integrating context-aware features and performing a feature selection considering a deeper analysis.

  1. Abdul Aziz et al., CSECU-DSG at CheckThat! 2023: Transformer-based Fusion Approach for Multimodal and Multigenre Check-Worthiness, CLEF 2023: Conference and Labs of the Evaluation Forum, September 18–21, 2023, Thessaloniki, Greece., 2023.

Causal Information Extraction

I have made notable contributions through published papers in areas such as causal event classification and the extraction of cause-effect relationships. I am particularly proud of achieving the top position twice in the Cause-Effect Relation Extraction shared task, reflecting my commitment to advancing our understanding of causality in natural language and its applications, with the ultimate aim of unraveling the intricate web of causal relationships embedded within textual data. Expanding my work in Causal Information Extraction to address real-world problems holds immense potential for practical applications. The ability to extract causal relationships from textual data can have far-reaching impacts in fields such as healthcare, finance, and environmental science, where understanding cause-and-effect dynamics is critical for decision-making and problem-solving. By leveraging my expertise in causal event classification and cause-effect relation extraction, I can contribute to the development of data-driven solutions that enhance our ability to analyze and predict causal factors in complex systems, ultimately leading to more informed and effective strategies for addressing real-world challenges. My research has the potential to bridge the gap between theoretical advancements and tangible benefits for society.

    li>Akram Hossain and Abdul Aziz et al., BiGCAT: A Hybrid Graph Neural Network-Based Approach to Named Entity Recognition for Financial Text, NAACL 2025 [Submitted].
  1. Abdul Aziz et al., CSECU-DSG@ Causal News Corpus 2022: Fusion of RoBERTa Transformers Variants for Causal Event Classification, Proceedings of the 5th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text(CASE), CASE@EMNLP '22, 2022.
  2. Abdul Aziz et al., Feature Fusion with Hand-crafted and Transfer Learning Embeddings for Cause-Effect Relation Extraction Feature Fusion with Hand-crafted and Transfer Learning Embeddings for Cause-Effect Relation Extraction, CEREX@FIRE 2020, Forum for Information Retrieval Evaluation (FIRE '20), 2020.
  3. Akram Hossain and Abdul Aziz et al., CSECU-DSG @Causal News Corpus 2023: Leveraging RoBERTa and DeBERTa Transformer Model with Contrastive Learning for Causal Event Classification, Proceedings of the 6th Workshop on Challenges and Applications of Automated Extraction of Socio-political Events from Text (CASE), CASE@RANLP '2023, 2023.

Biomedical and Health Informatics

My research interest in Biomedical and Health Informatics has evolved through my contributions to abstract text classification, where I achieved the second position in a shared task. Building upon this foundation, I aspire to broaden my impact in the field by exploring diverse facets such as genome sequencing, automatic clinical report generation, biomedical time series data analysis, Clinical Document summarization, bioNER (Biomedical Named Entity Recognition), biomedical Question Answering, and the development of large-scale multimodal models for biomedical data analysis. By delving into these areas, I aim to advance our understanding of complex biomedical data, foster innovation in healthcare, and ultimately contribute to the development of cutting-edge solutions that can improve patient care and drive biomedical research forward.

  1. Abdul Aziz et al., Enhancing the DeBERTa Transformers Model for Classifying Sentences from Biomedical Abstracts, Proceedings of the The 20th Annual Workshop of the Australasian Language Technology Association (ALTA '22), 2022.

Human Value Identification

My research passion lies in the intricate domain of human value identification, where I have already made contributions through published papers. My future ambitions in this field include the development of large-scale models, aiming to harness the power of artificial intelligence to better understand, identify, and analyze human values across diverse contexts and populations. By extending my work in this direction, I aspire to create comprehensive frameworks that can illuminate the intricate tapestry of human values, ultimately facilitating more informed decision-making and fostering greater empathy and understanding in our rapidly evolving world.

  1. Abdul Aziz et al., CSECU-DSG at SemEval-2023 Task 4: Fine-tuning DeBERTa Transformer Model with Cross-fold Training and Multi-sample Dropout for Human Values Identification, Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval '23), 2023.

Undergrad Thesis

During my undergrad thesis, I studied the multimodal learning paradigm. I put my knowledge to solve multimodal human desire understanding task. Recently, I have learned multitask learning and contrastive learning.

  1. Abdul Aziz et al., MMTF-DES: A Fusion of Multimodal Transformer Models for Desire, Emotion, and Sentiment Analysis of Social Media Data, Information Fusion Journal , Vol. -, No. -, pp. -, 2023. (On Review).