WG2 focuses on efficient ML machinery behind state-of-the-art Language Generation (LG) models, since much of the improvements we observe across different LG tasks nowadays are due to the application of varied deep neural network architectures. This involves, among others:
- multi-task learning,
- transfer learning,
- representation learning,
- structured prediction
- generative models
Moreover, WG2 investigates integration strategies for multi-modal data, which is of critical importance for Multi3Generation.
This working group will be working on:
- a repository of open source software for processing language and visual content, inc. a directory of sources of materials or components
- a survey on efficient use of ML for LG
- the organization of training schools
If you would like to join this working group, please contact the WG Leader and coleader: Aykut Erdem (aykut-at-cs.hacettepe.edu.tr) and Elena Lloret (elloret-at-dlsi.ua.es)
Publications and preprints
Please include full references and links where available.
- Amac, M. Sercan, Yagcioglu, Semih, Erdem, Aykut, and Erdem, Erkut. “Procedural Reasoning Networks for Understanding Multimodal Procedures”, Proceeding of the Conference on Natural Language Learning. 2020. Link to paper
- Milewski, Victor, Moens, Marie-Francine and Calixto, Iacer. “Are scene graphs good enough to improve image captioning?”, Proceedings of the 1st Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 10th International Joint Conference on Natural Language Processing. 2020. Link to paper
- Ales Zamuda and Elena Lloret. “Optimizing Data-Driven Models for Summarization as Parallel Tasks”. Journal of Computational Science, Vol. 42, April 2020. Link to paper
|Dataset name and brief description, including purpose||Authors/creators||Link|
|Name and/or brief description||Authors/creators||Link|
|Procedural Reasoning Networks||Amac, M. Sercan, Yagcioglu, Semih, Erdem, Aykut, and Erdem, Erkut||https://github.com/hucvl/prn|
Open source repository
Add anything here that doesn’t clearly fall under the other headings.