Hyewon Jang

About Me

I am a PhD student in computational psycholinguistics at the University of Konstanz supervised by Diego Frassinelli and Bettina Braun. I use experimental and computational linguistics methods to investigate the pragmatic dimensions of language that make human language complex and fun, with sarcasm being the current topic of interest.

I got into computational psycholinguistics after studying Speech and Language Processing for my master's degree at the University of Konstanz. Before choosing an academic path, I was an avid learner and teacher of languages — I received a bachelor's degree in English Education at Seoul National University and worked as an ESL teacher for a little while after that.

Research Interests

  • Experimental Linguistics

I test hypotheses about language phenomena by designing controlled online experiments for human participants. I also use these experiments to collect data for computational modeling experiments.

  • Computational Linguistics

I experiment with large language models to probe into what they can and cannot do. I enjoy comparing model behavior with human behavior as a useful pointer.

  • Multimodality

Language is the result of communications that happen in multiple modalities. My experiments involve textual, prosodic, and visual modalities for more comprehensive results about language and communication.

  • Figurative Language

I investigate figurative language, such as sarcasm, irony, or humor using the above methodologies.

Peer-Reviewed Publications

  (to appear) Hyewon Jang & Diego Frassinelli, Generalizable Sarcasm Detection is Just Around the Corner, of Course!, Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL 2024).

Summary
In this paper, we test the generalizability of sarcasm detection models by comparing three different language models finetuned on several datasets of sarcasm, including a new dataset we release (CSC), collected from multiple psycholinguistic experiments. We show that all language models finetuned on one dataset perform a lot worse on the other datasets, but that (CSC) can handle generalizable sarcasm detection relatively well. We discuss the reasons for the results in terms of the varied domains, styles, and label sources of sarcasm.

  (under review) Hyewon Jang, Bettina Braun, Diego Frassinelli, Contextual Factors that Trigger Sarcasm, Metaphor and Symbol .

Summary
In this paper, we discuss the contextual factors that trigger sarcasm based on three connected experiments. We confirm our hypothesis from these experiments that certain contextual factors motivate speakers to convey communicative functions traditionally associated with sarcasm, hence triggering the use of sarcasm. We also investigate interlocutor dynamics connected to the use of sarcasm.

  Hyewon Jang*, Qi Yu*, Diego Frassinelli, Figurative Language Processing: A Linguistically Informed Feature Analysis of the Behavior of Language Models and Humans, Findings of the Association for Computational Linguistics: ACL 2023 (Findings @ ACL 2023).
    *equal contribution

Summary
In this paper, we investigate what happens under the hood when Transformer models and traditional machine learning models perform figurative language classification (sarcasm, simile, idiom, and metaphor). We perform feature analyses on these models to compare general behaviors of different models, and to identify general tendencies of all models for different types of figurative language. We further compare model behavior with human behavior and provide insight into the different levels of complexity for different types of figurative language.

  Hyewon Jang, Bettina Braun, Diego Frassinelli, Intended and Perceived Sarcasm Between Close Friends: What Triggers Sarcasm and What Gets Conveyed?, Proceedings of the 45th Annual Conference of the Cognitive Science Society (CogSci 2023).

Summary
In this paper, we investigate what factors trigger sarcasm between close friends and whether the intentions behind a sarcastic comment are also conveyed to the listener. We answer our research questions using two connected experiments and analyze the results with the linear mixed-effects model.

Other Publications

  Hyewon Jang, Lexicon-Based Profiling of Irony and Stereotype Spreaders, Conference and Labs of the Evaluation Forum (CLEF 2022).

Summary
In this working note, I report the results of my experiments on profiling of Twitter authors that spread irony and stereotypes. I also report results from an ablation experiment to identify some of the most informative features for irony and stereotype profiling. This work is the result of my participation in The PAN 22 Author Profiling Shared Task (IROSTEREO).

  Ana-Maria Bucur, Hyewon Jang, Farhana Ferdousi Liza, Capturing Changes in Mood Over Time in Longitudinal Data Using Ensemble Methodologies, In Proceedings of the Eighth Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2022), pages 205–212, Seattle, USA. Association for Computational Linguistics.

Summary
In this working note, we describe our system that identifies changes in mood and behavior in longitudinal text data. This work is the result of our participation in the CLPsych 2022 Shared Task.