DiMLex: A Lexicon of Discourse Markers for Text Generation and Understanding
Summary
Discourse markers such as because or although indicate which coherence relation connects adjacent text spans. NLP systems have usually treated them as grammatical function words and handled them as a by-product of syntax. Stede and Umbach argue that they need lexicon-level treatment. Near-equivalent markers for one relation differ in meaning, presupposition, style and syntactic behaviour, and they show familiar lexical relations such as synonymy, antonymy, hyponymy and polysemy.
This position paper presents the design of DiMLex, a discourse marker lexicon that starts with German and plans to add English. Entries combine "top-down" insights from text linguistics with "bottom-up" descriptions from grammars, dictionaries and studies of individual markers. The authors outline a tentative feature inventory and two applications. In generation, marker selection becomes part of lexical choice. In understanding, the lexicon supplies cue lists and disambiguation hints for rhetorical parsing and dialogue processing.
The paper reports no evaluation. Its contribution is a resource design that prepares the ground for meaning-driven handling of discourse markers.
Key points
- Tentative feature inventory. The features cover:
- part of speech and positional freedom, and conjunct order;
- negation requirements (German sondern) and tense/aspect constraints;
- semantic class and polarity (if/unless) and commentability (weil vs. denn);
- speaker intention (aber vs. sondern) and presupposition (since vs. because);
- illocution type, stylistic register and the expressed discourse relation.
- Theory and formalism. RST serves as inspiration but is not adopted wholesale. An inheritance-based formalism is envisioned to cover hyponymy and polysemy.
- Generation view. Lookup starts from the discourse relation to be expressed. Information is grouped into applicability conditions, syntagmatic constraints, and preference criteria (e.g., stylistic) for choosing among the remaining candidates. Leaving a relation unmarked is also an option, and marker choice is integrated into sentence planning (Grote, Stede 1998).
- Understanding view. DiMLex can support cue-based rhetorical parsing in the style of Marcu (1997) through shallow structural tests. For example, German während allows only the temporal reading when followed by a noun phrase. It can also help disambiguate German discourse particles in dialogue, as studied in VERBMOBIL.
- Project status. The first phase runs until mid-1999 and focuses on German CONTRAST and CONCESSION markers. A generation testbed for paragraph-size text already exists. The planned outcome is a German/English generator with a prototype DiMLex for contrastive markers.
Cited by
- Constructing a Lexicon of English Discourse Connectives
- Inducing Discourse Connectives from Parallel Texts
- Multilingual Annotation and Disambiguation of Discourse Connectives for Machine Translation
- LEXCONN: A French Lexicon of Discourse Connectives
Cite
@inproceedings{stede1998dimlex,
title = {{DiMLex: A Lexicon of Discourse Markers for Text Generation and Understanding}},
author = {Stede, Manfred and Umbach, Carla},
booktitle = {36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics, Volume 2},
year = {1998},
doi = {10.3115/980691.980771},
url = {https://doi.org/10.3115/980691.980771}
}References
- N. Asher. Reference to abstract objects in Discourse. Dordrecht: Kluwer, 1993. find
- K. Barker. "Clause-level relationship analysis in the TANKA system." Technical report, Dept. of Computer Science, University of Ottawa, TR-94-07, 1994. find
- U. Brausse, E. Breindl-Hiller, R. Pasch. "Handbuch der deutschen Konnektoren." Institut fiir deutsche Sprache, Mannheim. In preparation. B. Di Eugenio, J. Moore, M. Paolucci. "Learning features that predict cue usage." In:. Proceedings of the 35th Annual Meeting of the ACL and 8th Conference of the… find
- B. Grote, M. Stede. "Discourse marker choice in sentence planning." To appear in: Proceedings of the 9th International Workshop on Natural Language Generation, Niagara-on-the-lake/Canada, 1998. find
- G. Helbig, J. Buscha. Deutsche Grammatik: Ein Handbuch f~r den AusMnderunterricht. Berlin, Leipzig: Langenscheidt, Verlag Enzyklop~.die, 1990. find
- A. Knott, C. Mellish. "A feature-based account of the relations signalled by sentence and clause connectives." In: Language and Speech 39 (2-3), 1996. find
- W. Mann, S. Thompson. "Rhetorical structure theory: Towards a functional theory of text organization." In: TEXT, 8:243-281, 1988 find
- D. Marcu. "The rhetorical parsing of natural language text." In: Proceedings of the 35th Annual Meeting of the ACL and 8th Conference of the European Chapter of the ACL, Madrid, July 1997. find
- J. Martin. English Text - System and Structure. Philadelphia/Amsterdam: John Benjamins, 1992. find
- C. Paris, K. Vander Linden, M. Fischer, A. Hartley, L. Pemberton, R. Power, D. Scott. "A support tool for writing multilingual instructions." In: Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence (IJCAI-95), Montreal, 1995. find
- R. Quirk, S. Greenbaum, G. Leech, J. Svartvik. A Grammar of Contemporary English. Harlow: Longman, 1992 (20th ed.) find
- T. Sanders, W. Spooren, L. Nordman. "Towards a taxonomy of coherence relations." In: Discourse Processes 15, 1992. find
- M. Stede, B. Schmitz. "Discourse particles and routine formulas in spoken language translation." In: Proceedings of the ACL/ELSNET Workshop on Spoken Language Translation, Madrid, 1997. find
- K. Vander Linden, J. Martin. "Expressing rhetorical relations in instructional text" In: Computational Linguistics 21(1):29-58, 1995. find
- L. Wanner, E. Hovy. "The HealthDoc sentence planner." In: Proceedings of the Eighth International Workshop on Natural Language Generation, Herstmonceux Castle, June 1996. find
This page in Markdown → https://md.research.elamur.ai/free/papers/stede-1998-dimlex-lexicon-discourse-markers-text/