publications

2025

1. NLP4DH

   AI Assistant for Socioeconomic Empowerment Using Federated Learning

   Nahedand Abdelgaber, Labiba Jahan, Nino Castellano, and 7 more authors

   In The 5th International Conference on Natural Language Processing for Digital Humanities, 2025

   Abs Bib

   Socioeconomic status (SES) reflects an individual’s standing in society, from a holistic set of factors including income, education level, and occupation. Identifying individuals in low-SES groups is crucial to ensuring they receive necessary support. However, many individuals may be hesitant to disclose their SES directly. This study introduces a federated learning-powered framework capable of verifying individuals’ SES levels through the analysis of their communications described in natural language. We propose to study language usage patterns among individuals from different SES groups using clustering and topic modeling techniques. An empirical study leveraging life narrative interviews demonstrates the effectiveness of our proposed approach.

   @inproceedings{NLP4DH,
     title = {{AI} Assistant for Socioeconomic Empowerment Using Federated Learning},
     author = {Abdelgaber, Nahedand and Jahan, Labiba and Castellano, Nino and Oltmanns, Joshua R. and Gupta, Mehak and Zhang, Jia and Pednekar, Akshay and Basavaraju, Ashish and Velazquez, Ian and Ma, Zerui},
     booktitle = {The 5th International Conference on Natural Language Processing for Digital Humanities},
     year = {2025},
     url = {https://openreview.net/forum?id=CgGEyoPBug}
   }

2. NCUR

   Recommender Systems for University Curriculum Advising

   Zerui Ma, Michael Hahsler, and Peter Moore

   , Pittsburgh, PennsylvaniaFor my contribution to this research, check out this blog , Apr 2025

   Abs Bib

   Effective academic advising plays a crucial role in enhancing student success, yet universities face challenges in optimizing advising processes and course enrollment. To address these challenges, recommender systems have emerged as valuable tools for automating personalized academic guidance, due to its capability in deriving recommendations in a wide range of data with only a few user entries. Universities can optimize their course planning and help students make informed decisions for their academic path with recommender systems. This research develops a novel recommender system tailored to undergraduate students, leveraging data on curriculum requirements, prerequisite dependencies, and student preferences. By leveraging data on course offerings, curriculum requirements, prerequisite dependencies, student preferences, and other relevant factors, this study develops a recommender system that help universities increase student advising efficiencies and create more intuitive and student-centric curricula. We structured and processed complex curriculum data to create an algorithm-ready environment, simplifying the relationships between degree requirements and course offerings. This study evaluates multiple algorithms based on recommendation accuracy, computational efficiency, and their ability to meet degree requirements. The system ensures students fulfill all degree requirements while fostering academic engagement. By streamlining course selection and exploring possible degree paths, it helps students graduate on time and navigate complex curricula. This research also collects important metrics to accurately predict student enrollment for classes, enabling college departments to plan their course offerings better. The study poses significant benefit to university advising offices by reducing advisor workloads and encouraging student engagement, advancing academic achievement of the entire student body.

   @article{NCUR,
     title = {Recommender Systems for University Curriculum Advising},
     author = {Ma, Zerui and Hahsler, Michael and Moore, Peter},
     booktitle = {National Conference on Undergraduate Research},
     location = {Pittsburgh, Pennsylvania},
     volume = {},
     number = {},
     pages = {},
     year = {2025},
     month = apr,
     publisher = {},
   }

3. AAAI Spring Symposium

   A Recommender System Architecture for University Curriculum Advising

   Zerui Ma, Michael Hahsler, and Peter Moore

   In Association for the Advancement of Artificial Intelligence Spring Symposium, San Francisco, CaliforniaFor my contribution to this research, check out this blog , Mar 2025

   Abs Bib

   Effective academic advising plays a crucial role in student success, yet universities face challenges in optimizing advising processes and course enrollment. This task is complicated by the fact that several graduation requirements have to be met while also taking the students’ interests into account. Academic advising has historically been performed by a skilled human adviser. Universities can optimize course planning and help students make informed decisions about their academic path with recommender systems. This case study develops a goal-based agent recommender system based on a large language model tailored to undergraduate students, depending on curriculum requirements, prerequisite dependencies, and student preferences. The developed recommendation system helps universities increase student advising efficiency and create more intuitive and student-centric curricula. We show how to structure and process complex curriculum data to create an algorithm-ready environment, simplifying the relationships between degree requirements and course offerings. This study evaluates multiple algorithms based on recommendation accuracy, computational efficiency, and their ability to meet degree requirements while fostering academic engagement. By streamlining course selection and exploring possible degree paths, the system may also help students graduate on time and navigate complex curricula. This system also collects important metrics to accurately predict student enrollment for classes, enabling college departments to plan their course offerings better. The system poses a significant benefit to university advising offices by reducing advisor workloads and encouraging student engagement, advancing the academic achievement of the entire student body.

   @inproceedings{AAAI Spring Symposium,
     title = {A Recommender System Architecture for University Curriculum Advising},
     author = {Ma, Zerui and Hahsler, Michael and Moore, Peter},
     booktitle = {Association for the Advancement of Artificial Intelligence Spring Symposium},
     location = {San Francisco, California},
     volume = {},
     number = {},
     pages = {},
     year = {2025},
     month = mar,
     publisher = {},
   }

4. EMNLP 2025

   Under Review

   Zerui Ma, and Anonomyous Authors

   Feb 2025

   Bib

   @unpublished{EMNLP,
     title = {Under Review},
     author = {Ma, Zerui and Authors, Anonomyous},
     journal = {EMNLP},
     volume = {},
     number = {},
     pages = {},
     year = {2025},
     month = feb,
     publisher = {Natural Language Processing (NLP) techniques have the potential to transform personality assessments, moving beyond traditional closed-ended questionnaires to more detailed, language-based evaluations. However, Large Language Models (LLMs) are constrained by the number of tokens they can process, limiting their ability to effectively contextualize longer texts. In this study, we use lengthy life narrative texts from older adults, each exceeding 2000 tokens to assess the Five-Factor Model (FFM) personality traits: openness, conscientiousness, extraversion, agreeableness, and neuroticism. We employ a hierarchical transformer approach with stop-gradient technique, involving the fine-tuning of pre-trained language models with a sliding-window technique, followed by the use of Recurrent Neural Networks (RNNs) with attention layers to reduce the error from the pre-trained model and capture the full context. This hybrid approach benefits from meaningful contextual text embeddings from pre-trained models, while the RNNs effectively manage the long context — a limitation of language models. We compared with state-of-the-art LLMs with extended token limits, such as LLaMA and Longformer. While these models can process longer texts, our proposed hierarchical method demonstrated superior performance in both accuracy and efficiency. We also analyze attention weights from RNN attention layers through different interpretability techniques to validate the model's results.}
   }

2024

1. APA

   Language-Based AI Assessment for Personality Disorder Science and Practice (Under Review)

   Zerui Ma, and Anonomyous Authors

   , For parts I have been editing, check out this blog , Oct 2024

   Abs Bib

   Personality disorders (PDs) are at a crossroads in classification and conceptualization. Advances in artificial intelligence (AI) and natural language processing offer a promising avenue towards clarifying differences in PD models and improving research methodology, understanding, and ultimately clinical utility. The present study aims to advance the use of language for modeling and assessing PD. A representative sample of N = 1,405 older adults in St. Louis (33% Black, 65% white) completed a life narrative interview, the Structured Interview for DSM-IV Personality (SIDP-IV), the NEO-Personality Inventory-Revised (NEO-PI-R), and self-report measures of physical functioning and depressive symptoms. Language associated with personality was modeled in three ways: 1) Parameters from the RoBERTa language model were fine-tuned on the life narrative transcripts and personality assessments, 2) Topics in the life narratives were modeled with BERTopic, and 3) Psychological, structural, and cognitive features of language were extracted using Linguistic Inquiry and Word Count (LIWC). Features from each language model were then combined in feed-forward neural networks to create language models in each domain, which were then combined in a linear regression to create combined multimodal models for DSM-IV personality disorder and NEO-PI-R personality traits. Multimodal language models were the psychometrically validated through cross-sectional correlations with self-reported functioning measures. Findings demonstrate promise of language-based AI with exciting future opportunities to refine conceptual frameworks of PD and provide automatic personality assessment and prediction in the clinic.

   @unpublished{APA,
     title = {Language-Based AI Assessment for Personality Disorder Science and Practice (Under Review)},
     author = {Ma, Zerui and Authors, Anonomyous},
     journal = {Journal of Psychopathology and Clinical Science},
     location = {},
     volume = {},
     issue = {},
     pages = {},
     numpages = {},
     year = {2024},
     month = oct,
     publisher = {APA},
   }