Retrospective and Future Views
Paper · 2026 · Kathleen McKeown
Summary
This article is a lightly edited version of Kathleen McKeown's acceptance talk for the 2025 ACL Lifetime Achievement Award, given in Vienna. It is not an experimental paper. It follows four strands of her research from their early beginnings to present-day projects: producing natural language automatically, condensing documents into summaries, studying content on social platforms, and examining artwork through multimodal methods.
She treats generation as a sequence of decisions about syntax, wording and ordering. Early work on floating constraints for lexical choice, encoded in the functional unification formalism and studied on a basketball report corpus, is presented as a precursor of today's controllable generation. Her group now pursues this with text diffusion models and with perspective-aware generation trained via reinforcement learning.
For summarization, she describes how a rejected NSF proposal led to extraction, compression and sentence fusion methods, and eventually to the NewsBlaster system.
Video statements from current students outline open challenges: evaluation, interpretability, bias and academia's role. The text is valuable as a record of research strategy, such as choosing unfashionable topics and persisting after rejection, and of mentoring lineage.
Key points
- Early ACL context: her first ACL Program Committee had about twenty members, each reviewing roughly 20–30 papers. Her ACL presidency coincided with the 1992 launch of the Linguistic Data Consortium.
- Summarization work began in 1996 with Dragomir Radev after the NSF called the first proposal infeasible. Two NSF grants followed a year later, funding cut-and-paste and text fusion methods for multi-document news.
- NewsBlaster went live on the Web right after 9/11 (2001). It ran daily, gathering articles from many news sites and clustering them by event.
- Current control research includes:
- style transfer along a formal–informal continuum with text diffusion models (Horvitz et al. 2024a,b);
- a general control framework built with NYU's Singhal and Ranganath (Singhal et al. 2025);
- a multi-country news corpus for perspective-based generation (Zhang, McKeown, and Muresan 2025).
- On mentoring, she names her first graduate Cecile Paris (CSIRO) and alumni such as Regina Barzilay, Michael and Noemie Elhadad, Min-Yen Kan and Ani Nenkova. She also lists alumni now at Anthropic, Amazon and Nvidia.
Institutions
United StatesFields, topics and concepts
📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksGeneral Social Sciences · 227 works
Cite
Cite as Kathleen McKeown. 2026. Retrospective and Future Views. Computational Linguistics.
@article{mckeown2026retrospective,
title = {{Retrospective and Future Views}},
author = {Kathleen McKeown},
journal = {Computational Linguistics},
year = {2026},
doi = {10.1162/coli.a.605},
url = {https://doi.org/10.1162/coli.a.605}
}References
- Alshomary, Milad, Narutatsu Ri, Marianna Apidianaki, Ajay Patel, Smaranda Muresan, and Kathleen McKeown. 2025. Latent space interpretation for stylistic analysis and explainable authorship attribution. In Proceedings of the 31st International Conference on Computational Linguistics, pages 1124–1135. find
- Ananthram, Amith, Elias Stengel-Eskin, Mohit Bansal, and Kathleen McKeown. 2025. See it from my perspective: How language affects cultural bias in image understanding. In The Thirteenth International Conference on Learning Representations. find
- Ananthram, Amith, Elias Stengel-Eskin, Lorena A. Bradford, Julia Demarest, Adam Purvis, Keith Krut, Robert Stein, Rina Elster Pantalony, Mohit Bansal, and Kathleen McKeown. 2026. PoSH: Using scene graphs to guide LLMs-as-a-judge for detailed image descriptions. In The Fourteenth International Confer… find
- Bin, Yi, Wenhao Shi, Yujuan Ding, Zhiqiang Hu, Zheng Wang, Yang Yang, See-Kiong Ng, and Heng Tao Shen. 2024. GalleryGPT: Analyzing paintings with large multimodal models. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 7734–7743. https://doi.org/10.1145/3664647.3681656… — [doi:10.1145/3664647.3681656](https://doi.org/10.1145/3664647.3681656) ↗
- Chang, Serina, Ruiqi Zhong, Ethan Adams, Fei-Tzin Lee, Siddharth Varia, Desmond Patton, William Frey, Chris Kedzie, and Kathy McKeown. 2018. Detecting gang-involved escalation on social media using context. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pa… find
- Deas, Nicholas and Kathleen McKeown. 2025. Summarization of opinionated political documents with varied perspectives. In Proceedings of the 31st International Conference on Computational Linguistics, pages 8088–8108. find
- Deas, Nicholas, Elsbeth Turcan, Ivan Ernesto Perez Mejia, and Kathleen McKeown. 2024b. MASIVE: Open-ended affective state identification in English and Spanish. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20467–20485. https://doi.org/10.18653/v1 /… find
- Subbiah, Melanie, Faisal Ladhak, Akankshya Mishra, Griffin Thomas Adams, Lydia Chilton, and Kathleen McKeown. 2024a. STORYSUMM: Evaluating faithfulness in story summarization. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 9988–10005. https://doi.org… ↗
- Turcan, Elsbeth. 2024. Detecting and explaining emotional reactions in personal narrative. Ph.D. thesis. https://doi.org/10 .7916/FXXX-WX64 Zhang, Tianyi, Faisal Ladhak, Esin Durmus, Percy Liang, Kathleen McKeown, and Tatsunori B. Hashimoto. 2024. Benchmarking large language models for news summariz…
Updated
This page in Markdown → https://md.research.elamur.ai/free/en/papers/mckeown-2026-retrospective-future-views/