Elamur Research⌕
☰
home / papers / 2024 / Discourse Structure for the Minecraft Corpus

Discourse Structure for the Minecraft Corpus

Paper · 2024 · Kate Thompson, Julie Hunter and Nicholas Asher

Summary

The paper presents the Minecraft Structured Dialogue Corpus (MSDC), a discourse-annotated layer over the Minecraft Dialogue Corpus (MDC). In MDC, an Architect guides a Builder through chat to construct a 3D object. Earlier work on turning these chats into building actions scored poorly, with verified F1 around 0.2. The authors argue that one instruction often takes several turns to settle: clarifications, questions, corrections and confirmations, some of them expressed through actions rather than words.

The team annotated nearly every dialogue with full SDRT graphs. These graphs combine linguistic units (EDUs) with nonlinguistic Builder moves (EEUs) and allow a unit to have several parents. To show the structures can be learned, they trained a discourse parser with a two-pass architecture. It performs well on both attachment and relation labeling, and especially on long-distance links.

The resource matters for situated, instruction-following agents such as cobots. It exposes how a task is broken down: negotiation episodes chained by long Narration links. This differs from chat corpora like STAC.

Key points

  • Coverage:
    • MSDC covers 541 of the 547 MDC dialogues, with 6 problematic games left out.
    • It keeps the original train/test split used by Jayannavar et al. (2020).
  • Corpus size:
    • Roughly 22.5k EDUs and about 34.6k relation instances.
    • About 6.3k discourse units have multiple parents.
  • Relations:
    • The inventory has 16 relation types.
    • The standard SDRT set is used minus Background and Parallel, plus a new Confirmation Question relation for Builder checks like "like that?".
  • Annotation process:
    • Three linguists and two NLP experts annotated in GLOZZ.
    • Each annotation was reviewed twice by two other linguists.
    • Scripts flagged further annotation errors.
  • Structural findings:
    • Narration carries more long-distance relations than any other relation type.
    • It forms the macrostructure linking negotiation episodes, which is unlike STAC, where Narration is rare and short-range.

Interesting ideas

  • I1. Discourse structure is modeled as a directed acyclic graph with a single head rather than a tree, so one reply ("not quite") can both answer a question and correct an earlier action.
  • I2. Physical actions are treated as discourse units (EEUs) that words can question, confirm or correct, putting events and utterances in one graph.
  • I3. Long runs of trivially related action moves are "squished" into one unit so that predictable Sequence links do not drown out rarer, more informative relations.
  • I4. Dialogue is analyzed as episodes (instruction → negotiation → acknowledgment) chained by Narration, giving a backbone of how a task was decomposed.
  • I5. Relation statistics are reported by distance between arguments, which shows where long-range attachment is hard.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): Let a speech-act or discourse node in prag attach to several antecedents, so a single utterance can carry answer and correction roles at once.
  • I2 (world): Represent narrated events and dialogue moves in the same world graph, so a character's reply can be linked to the action it comments on.
  • I3 (learn): When building silver training data, collapse highly regular chains (e.g., repeated steps) so that rare labels get enough weight during training.
  • I4 (world): Segment fables or word problems into episodes joined by narrative links, which supports event ordering and "what happened next" questions.
  • I5 (learn): Break down parser and coreference evaluation by arc or link distance, so gains on long-range cases are not hidden by easy local ones.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 5The 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 worksLiterature and Literary Theory · 519 works

🧭 Topics 1Research subjects and languages the work is about.Discourse analysis · 466 works

Cite

Cite as Kate Thompson, Julie Hunter and Nicholas Asher. 2024. Discourse Structure for the Minecraft Corpus. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{thompson2024discourse,
  title = {{Discourse Structure for the Minecraft Corpus}},
  author = {Kate Thompson, Julie Hunter and Nicholas Asher},
  booktitle = {Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
  year = {2024},
  doi = {10.63317/2hf5qt7c9iwj},
  url = {https://doi.org/10.63317/2hf5qt7c9iwj}
}

References

30 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Nicholas Asher. 1993. Reference to abstract objects in discourse, volume 50. Springer Science & Business Media. find
  2. Nicholas Asher, Julie Hunter, Mathieu Morey, Farah Benamara, and Stergos D Afantenos. 2016. Discourse structure and dialogue acts in multiparty dialogue: the STAC corpus. In LREC.
  3. Nicholas Asher, Julie Hunter, and Kate Thompson. 2020. Modelling structures for situated discourse. Dialogue & Discourse, 11:89–121. ↗
  4. Nicholas Asher and Alex Lascarides. 2003. Logics of conversation. Cambridge University Press. find
  5. Laura Banarescu, Claire Bonial, Shu Cai, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Kevin Knight, Philipp Koehn, Martha Palmer, and Nathan Schneider. 2013. Abstract meaning representation for sembanking. In Proceedings of the 7th linguistic annotation workshop and interoperability with discourse, pages 178–186.
  6. Zineb Bennis, Julie Hunter, and Nicholas Asher. 2023. A simple but effective model for attachment in discourse parsing with multi-task learning for relation labeling. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 3404–3409. find
  7. Claire Bonial, Mitchell Abrams, David Traum, and Clare Voss. 2021. Builder, we have done it: evaluating & extending dialogue-AMR NLU pipeline for two collaborative domains. In Proceedings of the 14th International Conference on Computational Semantics (IWCS), pages 173–183. find
  8. Claire Bonial, Lucia Donatelli, Mitchell Abrams, Stephanie Lukin, Stephen Tratz, Matthew Marge, Ron Artstein, David Traum, and Clare Voss. 2020. Dialogue-AMR: abstract meaning representation for dialogue. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 684–695. find
  9. Julia Bonn, Martha Palmer, Jon Cai, and Kristin Wright-Bettner. 2020. Spatial AMR: Expanded spatial annotation in the context of a grounded minecraft corpus. In Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020),. find
  10. Barbara G Deutsch, Grosz. 1974. The structure of task oriented dialogs. Artificial Intelligence Center, SRI International. find
  11. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. BERT: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805. ↗
  12. K. Forbes, E. Miltsakaki, R. Prasad, A. Sarkar, A. K. Joshi, and B. L. Webber. 2003. D-LTAG system: Discourse parsing with a lexicalized tree-adjoining grammar. Journal of Logic, Language and Information, 12(3):261–279. ↗
  13. Barbara J Grosz and Candace L Sidner. 1986. Attention, intentions, and the structure of discourse. Computational linguistics, 12(3):175–204. ↗
  14. Julie Hunter, Nicholas Asher, and Alex Lascarides. 2018. A formal semantics for situated conversation. Semantics and Pragmatics, 11. DOI: http://dx.doi.org/10.3765/sp.11.10. ↗
  15. Prashant Jayannavar, Anjali Narayan-Chen, and Julia Hockenmaier. 2020. Learning to execute instructions in a Minecraft dialogue. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2589–2602, Online. Association for Computational Linguistics. find
  16. Jiaqi Li, Ming Liu, Min-Yen Kan, Zihao Zheng, Zekun Wang, Wenqiang Lei, Ting Liu, and Bing Qin. 2020. Molweni: A challenge multiparty dialogues-based machine reading comprehension dataset with discourse structure. In Proceedings of the 28th International Conference on Computational Linguistics, pages 2642–2652.
  17. W. Mann and S. Thompson. 1987. Rhetorical structure theory : a theory of text organization. Technical report, Information Science Institute. find
  18. Y. Mathet and A. Widlöcher. 2009. La plate–forme GLOZZ : environnement d’annotation et d’exploration de corpus. In Traitement automatique des langues naturelles. find
  19. Mathieu Morey, Philippe Muller, and Nicholas Asher. 2018. A dependency perspective on RST discourse parsing and evaluation. Computational Linguistics, 44(2):197–235. ↗
  20. Anjali Narayan-Chen, Prashant Jayannavar, and Julia Hockenmaier. 2019. Collaborative dialogue in Minecraft. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5405–5415, Florence, Italy. Association for Computational Linguistics.
  21. Jérémy Perret, Stergos Afantenos, Nicholas Asher, and Mathieu Morey. 2016. Integer linear programming for discourse parsing. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 99–109, San Diego, California. ACL.
  22. L. Polanyi, C. Culy, M. van den Berg, G. L. Thione, and D. Ahn. 2004. A rule based approach to discourse parsing. In Proceedings of the 5th SIGdial Workshop on Discourse and Dialogue, pages 108–117, Cambridge, Massachusetts, USA. Association for Computational Linguistics.
  23. R. Prasad, N. Dinesh, A. Lee, E. Miltsakaki, L. Robaldo, A. Joshi, and B. Webber. 2008. The Penn Discourse Treebank 2.0. In Proceedings of LREC’08, Marrakech, Morocco. ELRA. Http://www.lrec-conf.org/pro-ceedings/lrec2008/.
  24. Zhengxiang Shi, Yue Feng, and Aldo Lipani. 2022. Learning to execute actions or ask clarification questions. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 2060–2070. find
  25. Zhouxing Shi and Minlie Huang. 2019. A deep sequential model for discourse parsing on multi-party dialogues. In AAAI. find
  26. Rajen Subba and Barbara Di Eugenio. 2009. An effective discourse parser that uses rich linguistic information. In Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 566–574.
  27. Ante Wang, Linfeng Song, Hui Jiang, Shaopeng Lai, Junfeng Yao, Min Zhang, and Jinsong Su. 2021. A structure self-aware model for discourse parsing on multi-party dialogues. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21, pages 3943–3949. International Joint Conferences on Artificial Intelligence Organization. Main Track. find
  28. Bonnie Webber, Norman Badler, Barbara Di Eugenio, Chris Geib, Libby Levison, and Michael Moore. 1995. Instructions, intentions and expectations. Artificial Intelligence, 73(1-2):253–269. ↗
  29. F. Wolf and E. Gibson. 2005. Representing discourse coherence: A corpus based study. Computational Linguistics, 31(2):249–287. ↗
  30. Wenhao Yu, Nimrod Gileadi, Chuyuan Fu, Sean Kirmani, Kuang-Huei Lee, Montse Gonzalez Arenas, Hao-Tien Lewis Chiang, Tom Erez, Leonard Hasenclever, Jan Humplik, Brian Ichter, Ted Xiao, Peng Xu, Andy Zeng, Tingnan Zhang, Nicolas Heess, Dorsa Sadigh, Jie Tan, Yuval Tassa, and Fei Xia. 2023. Language to rewards for robotic skill synthesis. Arxiv preprint arXiv:2306.08647. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/thompson-2024-discourse-structure-minecraft-corpus/