Maryam Hasan

PhD, Machine Learning, Deep Learning, Natural Language Processing

I am passionate about developing efficient machine learning and deep learning algorithms to solve real-world challenges in natural language processing (NLP) and natural language understanding (NLU). I completed my PhD in Computer Science at Worcester Polytechnic Institute (WPI) in June 2021 under the guidance of Professor Elke Rundensteiner and Professor Emmanuel Agu. My doctoral research focused on text classification using machine learning, deep transfer learning, and neural NLP techniques, including emotion detection and embedding learning via fine-tuned transformer architectures. Before that, I received my Master in Computer Science from University of Alberta in Canada.

Currently, I work as a Senior Data Scientist at Walmart Global Tech, where I apply and advance cutting-edge NLP methods in production. My recent work includes fine-tuning large language models (LLMs) for responsible AI in product safety, building retrieval-augmented generation (RAG)-based conversational agents for customer support, training deep learning models for product comparison and QA, and developing robust RoBERTa/BERT-based models for intent detection and entity recognition in e-commerce search-to-cart pipelines.

My ongoing research focuses on improving model generalization with limited labeled data through transfer learning and semi-supervised learning, with a strong emphasis on building scalable, safe, and customer-centric AI systems.

Google Scholar

CV

Projects

Responsible AI for Product Safety:

  • fine-tuned and evaluated Llama 3.2 (3B/1B) models using synthetically generated product Q&A data, reducing hallucination and harmful outputs in LLM responses.

    LLM-based Conversational Agents:

  • Built conversational AI agents using LLMs powered by retrieval-augmented generation (RAG) for customer support and shopping discovery.

    Product Comparison QA System:

  • Collaborated in designing and training a deep learning model to answer questions about several products and automate product comparison and Q&A resolution, accelerating decision-making.

Search-to-Cart NLP Modeling:

  • Developed and evaluated RoBERTa/BERT classifiers intent detection and entity recognition in shopping-related queries to improve e-commerce conversion pipelines.

DeepEmotex: A Deep Transfer Learning framework to Classify Emotion in Text Messages, 2021

  • Developed a deep transfer learning method to learn domain-specific features from context.
  • Fine-tuned pre-trained natural language models (e.g., BERT and USE) on the target classification task.
  • Developed a baseline neural network model (i.e., Bi-LSTM) and evaluated DeepEmotex models.
  • The proposed DeepEmotex-BERT model outperformed the baseline model by 23%.
  • Implemented using Python: Scikit-learn, NumPy, Pandas, TensorFlow, PyTorch. PDF
  • Developed a binary classification model to classify text messages into emotion and no-emotion classes.
  • Developed and evaluated an online method to measure public emotion and detect temporal changes of emotion in a stream of messages during events.
  • Used Hoeffding’s inequality to define an upper bound on the probability that the sum of independent random variables deviates from its expected value. PDF

Emotex: A Machine Learning approach to Detect Emotion in Text Streams, 2014

  • Collected and processed large corpus of labeled messages for supervised learning of emotions in text.
  • Developed and evaluated machine learning models to classify text messages including Support Vector Machines (SVM), Naïve Bayes and Decision Tree.
  • Implemented using Python (Scikit-learn, NumPy) and Java. PDF

LinCoh: A Feature Selection approach using Linear Coherent Bi-Clustering via Beam Searching and Sample Set Clustering, 2012

  • Developed a method to find linear coherent bi-clusters in Gene Expression Microarray data. Our method exploits a robust technique to identify conditionally correlated genes, combined with an efficient density-based search for clustering sample sets. PDF

NLP4SE: Analyzing Natural Language Artifacts of the Software Process, 2010

  • Designed and developed a tool to extract structured knowledge from textual data in software repositories (using Java, XQuery, DB2, XML, Stanford NLP tools: PoS tagger, Dependency Parser)
  • Extracted relations among named entities via Hierarchical Clustering of blogs (using Java, MySQL, Stanford NLP). PDF

Annoki, A Wiki tool for researchers, 2011

  • Collaborated in Annoki project which is a social Wiki tool for researchers
  • Developed a graphical web interface for Annoki
  • Implemented using Adobe Spring-Flex, PHP and MySQL

Publications

  • Maryam Hasan, Elke Rundensteiner, Emmanuel Agu, DeepEmotex: Classifying Emotion in Text Messages using Deep Transfer Learning, IEEE Big Data: Machine Learning on Big Data (IEEE BigData, MLBD), December 2021.
  • Maryam Hasan, Elke Rundensteiner, Emmanuel Agu, Automatic Emotion Detection in Text Streams by Analyzing Twitter data, International Journal of Data Science and Analytics, Springer 2019.
  • Maryam Hasan, Elke Rundensteiner, Xiangnan Kong, Using Social Sensing to Discover Trends in Public Emotion, In Proceedings of IEEE International Conference on Semantic Computing (IEEE ICSC), 2017.
  • Maryam Hasan, Elke Rundensteiner, Emmanuel Agu, Using Hashtags as Labels for Supervised Learning of Emotions in Twitter Messages, In Proceedings of ACM SIGKDD Workshop on Health Informatics (HI-KDD), August 2014.
  • Maryam Hasan, Elke Rundensteiner, Emmanuel Agu, EMOTEX: Detecting Emotions in Twitter Messages, In Proceedings of the 6th ASE International Conference on Social Computing, SocialCom, May 2014.
  • Di Yang, Kaiyu Zhao, Maryam Hasan, Hanyuan Lu, Elke Rundensteiner and Matthew Ward, Mining and Linking Patterns across Live Data Streams and Stream Archives, VLDB, October 2013.
  • Yi Shi, Maryam Hasan, Zhipeng Cai, Guohui Lin and Dale Schuurmans, Linear Coherent Bi-clustering via Beam Searching and Sample Set Clustering, International Journal of Discrete Mathematics, Algorithms and Applications (DMAA). December 2011.
  • Maryam Hasan, Eleni Stroulia, Denilson Barbosa and Manar Alalfi, Analyzing Natural Language Artifacts of the Software Process, Early Research Achievement track of the 26th IEEE International Conference on Software Maintenance (ICSM’2010), Timisoara, Romania, September 2010.
  • Yi Shi, Maryam Hasan, Zhipeng Cai, Guohui Lin and Dale Schuurmans, Linear coherent bi-cluster discovery via beam detection and sample set clustering, International Conference on Combinatorial Optimization and Applications (COCOA 2010). The Big Island, Hawaii, United States. December 2010.

Teaching

I had opportunities to teach several courses at SF Stae Universities to graduate and undergraduate students:

  • CSC775: Database Management Systems course to graduate students
  • CSC675: Database Management Systems course to undergraduate students
  • CSC146: Data Structures and Algorithms

I was Teaching Assistant for several courses at WPI:

  • CS 2102: Object Oriented Design using Java
  • DS 501, Introduction to Data Science
  • CS 4432, Database Systems II (Graduate students)
  • CS 5084, Algorithms design and analysis (Graduate students)