Ji-Ung Lee

Guest Lecture on Ethics in NLP

Fri, 23 Jan 2026 00:00:00 +0100

A guest lecture on ethics in NLP at the course computational linguistics.

Lecture Contents

Guest Lecture on Ethics in NLP

Fri, 24 Jan 2025 00:00:00 +0100

A guest lecture on ethics in NLP at the course computational linguistics.

Lecture Contents

Efficient Methods for Natural Language Processing: A Survey

Sat, 01 Jul 2023 00:00:00 +0200

Paper

Abstract

Recent work in natural language processing (NLP) has yielded appealing results from scaling model parameters and training data; however, using only scale to improve performance means that resource consumption also grows. Such resources include data, time, storage, or energy, all of which are naturally limited and unevenly distributed. This motivates research into efficient methods that require fewer resources to achieve similar results. This survey synthesizes and relates current methods and findings in efficient NLP. We aim to provide both guidance for conducting NLP under limited resources, and point towards promising research directions for developing more efficient methods.

Annotation Curricula to Implicitly Train Non-Expert Annotators

Fri, 01 Jul 2022 00:00:00 +0200

Abstract

Annotation studies often require annotators to familiarize themselves with the task, its annotation scheme, and the data domain. This can be overwhelming in the beginning, mentally taxing, and induce errors into the resulting annotations; especially in citizen science or crowd sourcing scenarios where domain expertise is not required and only annotation guidelines are provided. To alleviate these issues, we propose annotation curricula, a novel approach to implicitly train annotators. We gradually introduce annotators into the task by ordering instances that are annotated according to a learning curriculum. To do so, we first formalize annotation curricula for sentence- and paragraph-level annotation tasks, define an ordering strategy, and identify well-performing heuristics and interactively trained models on three existing English datasets. We then conduct a user study with 40 voluntary participants who are asked to identify the most fitting misconception for English tweets about the Covid-19 pandemic. Our results show that using a simple heuristic to order instances can already significantly reduce the total annotation time while preserving a high annotation quality. Annotation curricula thus can provide a novel way to improve data collection. To facilitate future research, we further share our code and data consisting of 2,400 annotations.

Bachelorpraktikum

Thu, 01 Oct 2020 00:00:00 +0200

Project coordinator of 50+ software projects for undergrads

Course Contents

The purpose of this course is to provide undergrads with hands-on experience in their first longer software project. For this, they are put together in teams of four to five students and receive a software project that are proposed by members of the university. These can for instance be to implement a user interface to visualize some cool research, or even programming a rasberry pie to automatically receive and manage orders for a coffee machine.

Empowering Active Learning to Jointly Optimize System and User Demands

Wed, 01 Jul 2020 00:00:00 +0200

Abstract

Existing approaches to active learning maximize the system performance by sampling unlabeled instances for annotation that yield the most efficient training. However, when active learning is integrated with an end-user application, this can lead to frustration for participating users, as they spend time labeling instances that they would not otherwise be interested in reading. In this paper, we propose a new active learning approach that jointly optimizes the seemingly counteracting objectives of the active learning system (training efficiently) and the user (receiving useful instances). We study our approach in an educational application, which particularly benefits from this technique as the system needs to rapidly learn to predict the appropriateness of an exercise to a particular user, while the users should receive only exercises that match their skills. We evaluate multiple learning strategies and user types with data from real users and find that our joint approach better satisfies both objectives when alternative methods lead to many unsuitable exercises for end users.

Bachelorpraktikum

Tue, 01 Oct 2019 00:00:00 +0200

Project coordinator of 50+ software projects for undergrads

Course Contents

Manipulating the Difficulty of C-Tests

Mon, 01 Jul 2019 00:00:00 +0200

Abstract

We propose two novel manipulation strategies for increasing and decreasing the difficulty of C-tests automatically. This is a crucial step towards generating learner-adaptive exercises for self-directed language learning and preparing language assessment tests. To reach the desired difficulty level, we manipulate the size and the distribution of gaps based on absolute and relative gap difficulty predictions. We evaluate our approach in corpus-based experiments and in a user study with 60 participants. We find that both strategies are able to generate C-tests with the desired difficulty level.

Text Analytics

Sun, 01 Apr 2018 00:00:00 +0200

Text analytics course on active learning for undergraduate and graduate students

Course Contents

Active Learning (not to confuse with Active Learning from the educational field) tackles the challenge of how machine learning algorithms can achieve an equal or even greater performance with less training data. Supervised learning methods require labeled data which is often costly to annotate and often involves experts. However, different data contributes differently to the performance of an machine learning algorithm. For example, when training a classifier to classify pictures of cats, dogs, and alligators, adding more cat pictures would not help that much when the classifier is already very good at recognizing cats. The main research question in active learning is which data gives us the biggest gain if they had labels. For this we search for different strategies which decrease the number of necessary training data, independent of the underlying machine learning algorithm.

Terms and Privacy Policy

Mon, 06 Jun 2016 00:00:00 +0200

Privacy Policy

The privacy of my visitors is extremely important. This Privacy Policy outlines the types of personal information that is received and collected and how it is used.

First and foremost, I will never share your email address or any other personal information to anyone without your direct consent.

Log Files

Like many other websites, this site uses log files to help learn about when, from where, and how often traffic flows to this site. The information in these log files include:

Community Engagement

Mon, 01 Jan 0001 00:00:00 +0000

Community Engagement

Since 2019, I am a regular reviewer at major NLP and machine learning conferences (please check my cv for details). I have received outstanding reviewer awards, most notably at ACL 2020, EACL 2021, and ARR. I also review at various workshops such as BEA.

CV

Mon, 01 Jan 0001 00:00:00 +0000

Download my CV (PDF).

Editing this website

Mon, 01 Jan 0001 00:00:00 +0000

This website is built with Hugo. Edit Markdown files in content/, site settings in hugo.toml, templates in layouts/, and styles in assets/css/main.css. See the project README for local preview and deployment instructions.

imprint

Mon, 01 Jan 0001 00:00:00 +0000

Impressum

Angaben gem. § 5 DDG

Betreiber und Kontakt: Ji-Ung Lee

Addresse: via E-Mail.

E-Mail: siehe Hauptseite.

Inhaltlich Verantwortlicher gem. § 55 II RStV

Ji-Ung Lee

Kontakt: siehe oben.

Datenschutz

Die Nutzung dieser Webseite ist ohne die Angabe von Daten möglich.

Haftungsausschluss

Die Inhalte dieser Webseite wurden mit großer Sorgfalt erstellt, erheben jedoch keinen Anspruch auf Richtigkeit oder Vollständigkeit. Sofern Links zu anderen Seiten bestehen, sind alleinig die dortigen Betreiber für den Inhalt verantwortlich. Für den Inhalt dieser Seiten wird keine Haftung übernommen.

Other Activities

Mon, 01 Jan 0001 00:00:00 +0000

No entries have been published here yet.

Page Archive

Mon, 01 Jan 0001 00:00:00 +0000

Portfolio

Mon, 01 Jan 0001 00:00:00 +0000

No entries have been published here yet.

Posts by Category

Mon, 01 Jan 0001 00:00:00 +0000

No entries have been published here yet.

Posts by Collection

Mon, 01 Jan 0001 00:00:00 +0000

Posts by Tags

Mon, 01 Jan 0001 00:00:00 +0000

No entries have been published here yet.

Publications

Mon, 01 Jan 0001 00:00:00 +0000

For an up-to-date list please also check my google scholar

Accepted Publications (latest first, ⋆ marks equal first/last author works.)

Anna Karnysheva, Dietrich Klakow⋆, Ji-Ung Lee⋆, Probing Chemical Language Models: Effects of Pre-training and Fine-tuning Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2026.
Sarubi Thillainathan, Ji-Ung Lee, Michael Sullivan, Alexander Koller, AuthorMix: Modular Authorship Style Transfer via Layer-wise Adapter Mixing Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2026.
Ji-Ung Lee, Marc Pfetsch, Iryna Gurevych. Constrainted C-Test Generation using Mixed-Integer Programming. arXiv:2404.08821. 2024.
Zena Al-Khalili, Rafi Hakim, Dietrich Klakow⋆, Ji-Ung Lee⋆, Fork-Think with Confidence, Conference on Language Modeling (COLM), 2026.
Yifan Wang, Mayank Jobanputra, Ji-Ung Lee, Soyoung Oh, Isabel Valera, Vera Demberg, Bridging Fairness and Explainability: Can Input-Based Explanations Promote Fairness in Hate Speech Detection? The Fourteenth International Conference on Learning Representations (ICLR), 2026.
Yifan Wang, Sukrut Rao, Ji-Ung Lee, Mayank Jobanputra, Vera Demberg, B-cos LM: Efficiently Transforming Pre-trained Language Models for Improved Explainability Transactions of Machine Learning Research (TMLR), 2025.
Marcos Treviso⋆, Ji-Ung Lee⋆, Tianchu Ji*, Betty van Aken, Qingqing Cao, Manuel R Ciosici, Michael Hassid, Kenneth Heafield, Sara Hooker, Pedro H Martins, André FT Martins, Peter Milder, Colin Raffel, Edwin Simpson, Noam Slonim, Niranjan Balasubramanian, Leon Derczynski, Roy Schwartz, Efficient Methods for Natural Language Processing: A Survey Transactions of the Association for Computational Linguistics, 11: pages 826–860. July, 2023. MIT Press
Jan-Christoph Klie, Ji-Ung Lee, Kevin Stowe, Gözde Gül Şahin, Nafise Sadat Moosavi, Luke Bates, Dominic Petrak, Richard Eckart De Castilho, Iryna Gurevych, Lessons Learned from a Citizen Science Project for Natural Language Processing Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 3594–3608, May, 2023. Dubrovnik, Croatia
Haishuo Fang, Ji-Ung Lee, Nafise Sadat Moosavi, and Iryna Gurevych, Transformers with Learnable Activation Functions, Findings of the Association for Computational Linguistics: EACL 2023, pages 2382–2398, May, 2023. Dubrovnik, Croatia
Lorenz Stangier⋆, Ji-Ung Lee⋆, Yuxi Wang, Marvin Müller, Nicholas Frick, Joachim Metternich, and Iryna Gurevych, TexPrax: A Messaging Application for Ethical, Real-time Data Collection and Annotation Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing: System Demonstrations (AACL), pages 9–16, November, 2022. Taipei, Taiwan.
Ji-Ung Lee⋆, Jan-Christoph Klie⋆, and Iryna Gurevych, Annotation Curricula to Implicitly Train Non-Expert Annotators Computational Linguistics, Volume 48 (2), pages 343–373, June, 2022. MIT Press
Tilman Beck, Ji-Ung Lee, Christina Viehmann, Marcus Maurer, Oliver Quiring, and Iryna Gurevych, Investigating label suggestions for opinion mining in German Covid-19 social media Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (ACL), pages 1–13, July, 2021. Online.
Marvin Müller, Ji-Ung Lee, Nicholas Frick, Lorenz Stangier, Iryna Gurevych, and Joachim Metternich, Extracting problem related entities from production chats to enhance the data base for assistance functions on the shop floor 9th CIRP Global Web Conference – Sustainable, resilient, and agile manufacturing and service operations : Lessons from COVID-19 (Procedia CIRP), Volume 103, pages 231–236, October, 2021. Online.
Marianne Grace Araneta, Gülşen Eryiğit, Alexander König, Ji-Ung Lee, Ana Luís, Verena Lyding, Lionel Nicolas, Christos Rodosthenous, Federico Sangati, Substituto–A Synchronous Educational Language Game for Simultaneous Teaching and Crowdsourcing In Proceedings of the 9th Workshop on NLP for Computer Assisted Language Learning (NLP4CALL) pages 1–9, November, 2020. Online.
Ji-Ung Lee, Christian M. Meyer, and Iryna Gurevych. Empowering Active Learning to Jointly Optimize System and User Demands Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), pages 4233–4247, July, 2020. Online
Ji-Ung Lee, Erik Schwan, and Christian M. Meyer, Manipulating the Difficulty of C-Tests Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics(ACL),pages 360–370, July, 2019. Florence, Italy
Steffen Eger, Gözde Gül Şahin, Andreas Rücklé, Ji-Ung Lee, Claudia Schulz, Mohsen Mesgar, Krishnkant Swarnkar, Edwin Simpson, Iryna Gurevych. Text Processing Like Humans Do: Visually Attacking and Shielding NLP Systems Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL-HLT), pages 1634–1647, June, 2019. Minneapolis, USA
Ji-Ung Lee, Steffen Eger, Johannes Daxenberger, and Iryna Gurevych, UKP TU-DA at GermEval 2017: Deep Learning for Aspect Based Sentiment Detection Proceedings of the GermEval 2017 - Shared Task on Aspect-based Sentiment in Social Media Customer Feedback, September, 2017. Berlin, Germany

Preprints (latest first, * marks equal contribution works.)

Yifan Wang, Jinyi Mu, Mayank Jobanputra, Yu Wang, Ji-Ung Lee, Soyoung Oh, Isabel Valera, Vera Demberg, Sparse Mixture-of-Experts Reward Models Learn Interpretable and Specialized Experts for Personalized Preference Modeling arXiv:2606.04284, 2026.
Ji-Ung Lee, Marc E. Pfetsch, Iryna Gurevych, Constrained c-test generation via mixed-integer programming arXiv:2404.08821, 2024.
Ji-Ung Lee, Haritz Puerto, Betty van Aken, Yuki Arase, Jessica Zosa Forde, Leon Derczynski, Andreas Rücklé, Iryna Gurevych, Roy Schwartz, Emma Strubell, Jesse Dodge. Surveying (Dis)Parities and Concerns of Compute Hungry NLP Research arXiv:2306.16900, 2023.
Ulf A Hamster, Ji-Ung Lee, Alexander Geyken, Iryna Gurevych. Rediscovering Hashed Random Projections for Efficient Quantization of Contextualized Sentence Embeddings arXiv:2304.02481, 2023.

Sitemap

Mon, 01 Jan 0001 00:00:00 +0000

Talks

Mon, 01 Jan 0001 00:00:00 +0000

Invited talk: Facets of efficiency in NLP (youtube). KUIS AI, Ko¸c University. September, 2022.

Investigating rational activation functions to train Transformer models

Spotlight talk: Investigating rational activation functions to train Transformer models. Dagstuhl Seminar on Efficient and Equitable Natural Language Processing in the Age of Deep Learning (webpage). June, 2022.

Talks and presentations

Mon, 01 Jan 0001 00:00:00 +0000

Invited talk: Facets of efficiency in NLP (youtube). KUIS AI, Ko¸c University. September, 2022.

Investigating rational activation functions to train Transformer models

Workshop Organization

Mon, 01 Jan 0001 00:00:00 +0000

Workshop Organization

Interactive learning for natural language processing (InterNLP).

InterNLP 2022 (co-located with NeurIPS 2022)
InterNLP 2021 (co-located with ACL 2021)