Unpacking Ambiguity: The Interaction of Polysemous Discourse Markers and Non-DM Signals

What it is about

Using GUM's eRST layer (16 genres, over 250K tokens, discourse markers plus seven types of non-marker signals: graphical, lexical, morphological, numerical, reference, semantic, syntactic), the paper measures how ambiguous each discourse marker is — the entropy of the relations it signals — and asks whether ambiguous markers come with more, or more varied, other signals. Overall so is the most polysemous marker; and is the most polysemous in six genres, so and as second in eight; for, and, if, by, when co-occur with the most diverse signals. Polysemy correlates with the diversity rather than the number of accompanying signals, and genre matters a lot: spoken genres (court, podcast, vlog) lean most on extra signals, and all 16 genres differ significantly from the overall signal distribution. The most frequent combination with and is reference + semantic (a shared referent plus related words across the two units); in letters as mostly signals manner and needs no extra signal.

Ideas

• I1. Compute the same entropy from Mova's own output: for each connective the reader handles, the distribution of link types it produced (or of teacher labels on the frame bench). High-entropy connectives (and, so, as, since, while, when) are where disambiguation rules pay off; low-entropy ones can stay single rules. A small report in prag/src/fbench.rs, recomputed per genre (tales, lessons, dialogue).
• I2. For ambiguous connectives, require a second signal of the kind the paper finds, each named in the explanation: and + shared referent → next (otherwise additive, no link); since + perfect tense or a time expression → temporal, else cause; as + a manner verb or as if → manner, + progressive/simultaneous action → temporal, else cause; while + contrasting subjects or antonyms → contrast, else temporal. Implement as tests in prag/src/frames.rs with PRAG_OFF switches; measure frame bench cause/next precision.
• I3. Keep genre as an input to these defaults (see scholman-2022-discogem-crowdsourced-corpus-genre-mixed I3); the paper shows the same marker shifts sense with genre (so, while, given, indeed).

For Mova

36. Discourse relations and coherence §6, §11.1, §11.6 (connective senses not disambiguated). Code: prag/src/frames.rs (second-signal tests), prag/src/fbench.rs (entropy report). Measures: frame bench cause/next precision per connective and per genre.

Topics: Discourse analysis · Word sense disambiguation

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.codi-1.2 · arXiv 2507.16748

@inproceedings{wu2025unpacking,
  title = {{Unpacking Ambiguity: The Interaction of Polysemous Discourse Markers and Non-DM Signals}},
  author = {Wu, Jingni and Zeldes, Amir},
  booktitle = {Proceedings of the 6th Workshop on Computational Approaches to Discourse, Context and Document-Level Inferences (CODI 2025)},
  year = {2025},
  doi = {10.18653/v1/2025.codi-1.2},
  eprint = {2507.16748},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.18653/v1/2025.codi-1.2}
}

References

28 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Cristina Grisot and Joanna Blochowiak. 2017. Temporal connectives and verbal tenses as processing instructions: Evidence from French. Pragmatics & Cognition, 24:404–440. find
2. Tatsuya Aoyama, Shabnam Behzad, Luke Gessler, Lauren Levine, Jessica Lin, Yang Janet Liu, Siyao Peng, Yilun Zhu, and Amir Zeldes. 2023. GENTLE: A genre-diverse multilayer challenge set for English NLP and linguistic evaluation. In Proceedings of the 17th Linguistic Annotation Workshop (LAW-XVII), pages 166–178, Toronto, Canada. Association for Computational Linguistics. ↗ our card
3. Jet Hoek, Sandrine Zufferey, Jacqueline Evers-Vermeul, and Ted JM Sanders. 2018. The linguistic marking of coherence relations: Interactions between connectives and segment-internal elements. Pragmatics & Cognition, 25(2):276–309. find
4. Fatemeh Torabi Asr and Vera Demberg. 2012. Measuring the strength of linguistic cues for discourse relations. Proceedings of the Workshop on Advances in Discourse Analysis and Its Computational Aspects, pages 33–42. find
5. Wei Liu, Yi Fan, and Michael Strube. 2023. HITS at DISRPT 2023: Discourse segmentation, connective detection, and relation classification. In Proceedings of the 3rd Shared Task on Discourse Relation Parsing and Treebanking (DISRPT 2023), pages 43–49, Toronto, Canada. The Association for Computational Linguistics. find
6. Chloé Braud, Amir Zeldes, Laura Rivière, Yang Janet Liu, Philippe Muller, Damien Sileo, and Tatsuya Aoyama. 2024. DISRPT: A multilingual, multi-domain, cross-framework benchmark for discourse processing. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 4990–5005, Torino, Italia. ELRA and ICCL. find
7. Yang Liu and Amir Zeldes. 2019. Discourse relations and signaling information: Anchoring discourse signals in RST-DT. In Proceedings of the Society for Computation in Linguistics (SCiL) 2019, volume 2, pages 314–317. find
8. William C. Mann and Sandra A. Thompson. 1988. Rhetorical Structure Theory: Toward a functional theory of text organization. Text, 8(3):243–281. ↗ our card
9. Herbert H Clark. 2002. Speaking in time. Speech Communication, 36(1-2):5–13. find
10. Rashmi Prasad, Bonnie Webber, and Alan Lee. 2018. Discourse annotation in the PDTB: The next generation. In Proceedings of the 14th Joint ACL-ISO Workshop on Interoperable Semantic Annotation, pages 87–97, Santa Fe, New Mexico, USA. Association for Computational Linguistics. find our card
11. Herbert H Clark and Jean E Fox Tree. 2002. Using uh and um in spontaneous speaking. Cognition, 84(1):73–111. find
12. Ludivine Crible. 2020. Weak and strong discourse markers in speech, chat, and writing: Do signals compensate for ambiguity in explicit relations? Discourse Processes, 57(9):793–807. find
13. Alan Robaldo, Eleni Miltsakaki, Alan Lee, Rashmi Prasad, Nikhil Dinesh, Bonnie Webber, and Aravind Joshi. 2008. The Penn Discourse Treebank 2.0. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), Marrakech, Morocco. ELRA. find our card
14. Ludivine Crible and Vera Demberg. 2020. The role of non-connective discourse cues and their interaction with connectives. Pragmatics & Cognition, 27(2):313–338. find
15. Debopam Das and Maite Taboada. 2018a. RST Signalling Corpus: A corpus of signals of coherence relations. Language Resources and Evaluation, 52:149– 184. find
16. Claude E Shannon. 1951. Prediction and entropy of printed English. Bell System Technical Journal, 30(1):50–64. find
17. Debopam Das and Maite Taboada. 2018b. Signalling of coherence relations in discourse, beyond discourse markers. Discourse Processes, 55(8):743–770. find
18. Maite Taboada and Debopam Das. 2013. Annotation upon annotation: Adding signalling information to a corpus of discourse relations. Dialogue & Discourse, 4(2):249–281. find
19. Debopam Das and Maite Taboada. 2019. Multiple signals of coherence relations. Discours, (24). find
20. Bonnie Webber, Rashmi Prasad, Alan Lee, and Aravind Joshi. 2019. The Penn Discourse Treebank 3.0 annotation manual. University of Pennsylvania, 35:108. find our card
21. Katherine Forbes-Riley, Bonnie Webber, and Aravind Joshi. 2006. Computing discourse semantics: The predicate-argument semantics of discourse connectives in D-LTAG. Journal of Semantics, 23(1):55– 106. find
22. Bonnie Lynn Webber and Aravind K Joshi. 1998. Anchoring a lexicalized tree-adjoining grammar for discourse. arXiv preprint cmp-lg/9806017. find
23. Luke Gessler, Siyao Peng, Yang Liu, Yilun Zhu, Shabnam Behzad, and Amir Zeldes. 2020. AMALGUM – a free, balanced, multilayer English web corpus. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 5267–5275, Marseille, France. European Language Resources Association. find our card
24. Yue Yu, Yilun Zhu, Yang Liu, Yan Liu, Siyao Peng, Mackenzie Gong, and Amir Zeldes. 2019. GumDrop at the DISRPT2019 shared task: A model stacking approach to discourse unit segmentation and connective detection. In Proceedings of the Workshop on Discourse Relation Parsing and Treebanking 2019, pages 133–143, Minneapolis, MN. Association for Computational Linguistics. find our card
25. Amir Zeldes. 2017. The GUM corpus: Creating multilayer resources in the classroom. Language Resources and Evaluation, 51(3):581–612. ↗
26. Amir Zeldes. 2022. Can we fix the scope for coreference? problems and solutions for benchmarks beyond OntoNotes. Dialogue & Discourse, 13(1):41– 62. ↗ our card
27. Amir Zeldes, Tatsuya Aoyama, Yang Janet Liu, Siyao Peng, Debopam Das, and Luke Gessler. 2025. eRST: A signaled graph theory of discourse relations and organization. Computational Linguistics, 51(1):23– 72. find our card
28. Amir Zeldes and Yang Janet Liu. 2020. A neural approach to discourse relation signal detection. Dialogue and Discourse, 11(2):74–99. Text type Source Academic writing Biographies CC Vlogs Conversations Courtroom transcripts Essays Fiction Forum How–to guides Interviews Letters News stories Podcasts Political speeches Textbooks Travel guides Various Wikipedia YouTube UCSB Corpus Various Various Various reddit wikiHow Wikinews Various Wikinews Various Various OpenStax Wikivoyage Total GUM Docs Tokens 18 20 15 15 9 9 19 18 19 19 12 24 10 15 15 18 17,169 18,213 16,864 17,932 11,148 10,842 17,511 16,364 17,081 18,196 9,989 17,186 11,986 16,720 16,693 16,515 250,409 *Table 6: Overview of GUM corpus by text type.* find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/wu-2025-unpacking-ambiguity-interaction-polysemous-discourse/