Elamur Research⌕
☰
home / papers / 2017 / Temporal@ODIL project: Adapting ISO-TimeML to syntactic treebanks for the temporal annotation of spoken speech

Temporal@ODIL project: Adapting ISO-TimeML to syntactic treebanks for the temporal annotation of spoken speech

Summary

The paper describes the Temporal@ODIL project, which aims to build the largest corpus of spoken French annotated with temporal information. Until now, the only French resource following the ISO-TimeML standard was the French Time Bank. That resource is small, is built on written texts, and does not take syntax into account. The authors annotate speech transcripts from the ANCOR corpus, which range from sociolinguistic interviews to lively dialogues. To do this, they adapt ISO-TimeML so that the extent of an event is a node of the syntactic tree rather than a minimal span. The annotator therefore just selects a single node in the treebank. The values of several attributes are also extended, while the standard's XML structure stays compatible. Annotation has five stages: automatic cleaning of the transcripts, syntactic parsing with the Stanford parser trained on the FTB, manual correction and deepening of the trees, and then manual annotation of events and temporal relations. A dedicated web tool was built for this work. Pilot experiments showed that this kind of annotation requires constituency trees rather than dependency trees. The result is an open resource that combines syntax, coreference and time for spoken French.

Key points

  • Resource size: a syntactic layer of 100,000 words and a temporal layer of 20,000 words. This is twice the size of existing French TimeML resources, since the French Time Bank has 15,876 words. An open release under the CC-BY-SA license was planned for June 2018.
  • The temporalFunction attribute has three values following Reichenbach: Null for absolute references, S for references anchored to the speech time, and R for references anchored to the discourse. The IDENTITY type in TLINK was dropped as purely coreferential, and several new relations were added instead.
  • On spontaneous French, dependency parsers showed no clear advantage over constituency parsers, so the Stanford parser trained on the FTB was chosen. Annotators also found it hard to determine extents on dependency trees.
  • The flat FTB trees are deepened by automatic rules. For example, a VN together with a VPinf, NP, PP or ADV in a given context is merged into a VP. Speaker self-corrections are annotated following the Rhapsodie guidelines.
  • The tool's client is written in Elm and runs in the browser. The server is written in Haskell, they communicate over websockets, and data is exchanged as JSON. The tool permits only operations that keep the trees well-formed, so it is also suitable for general treebank editing.

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: Linguistics & languages

Topics: Dependency parsing · Treebanks

Cite

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

@inproceedings{antoine2017temporal,
  title = {{Temporal@ODIL project: Adapting ISO-TimeML to syntactic treebanks for the temporal annotation of spoken speech}},
  author = {Antoine, Jean-Yves and Wasczuk, Jakub and Lefeuvre-Haftermeyer, Anaïs and Abouda, Lotfi and Schang, Emmanuel and Savary, Agata},
  booktitle = {Proceedings of the 13th Joint ISO-ACL Workshop on Interoperable Semantic Annotation (ISA-13)},
  year = {2017},
  url = {https://aclanthology.org/W17-7413/}
}

References

15 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Abeillé, A., L. Clément, and F. Toussenel (2003). Building a treebank for french. Treebanks, 165–187. find
  2. Allen, J. F. (1983). Maintaining knowledge about temporal intervals. Communications of the ACM 26. find
  3. Bejček, E. and P. Straňák (2010, Apr). Annotation of multiword expressions in the prague dependency treebank. Language Resources and Evaluation 44(1), 7–21. find
  4. Bittar, A., P. Amsili, and P. Denis (2011). French TimeBank : un corpus de référence sur la temporalité en français. In TALN 2011, Montpellier, France, pp. 259–270. find
  5. Drat, L. (2014). Projet TourInFlux. Annotation des expressions temporelles. Master’s thesis. find
  6. ISO (2012). Language resource management - Semantic annotation framework (SemAF) Part 1: Time and events. ISO 24617-1:2012, International Organization for Standardization, Geneva, Switzerland. find
  7. Lacheret, A., S. Kahane, J. Beliao, A. Dister, K. Gerdes, J.-P. Goldman, N. Obin, P. Pietrandrea, and A. Tchobanov (2014, July). Rhapsodie: un Treebank annoté pour l’étude de l’interface syntaxe-prosodie en français parlé. In 4e Congrès Mondial de Linguistique Française, Volume 8, Berlin, Germ… find
  8. Lefeuvre-Halftermeyer, A., J.-Y. Antoine, A. Couillault, E. Schang, L. Abouda, A. Savary, D. Maurel, I. Eshkol-Taravella, and D. Battistelli (2016). Covering various Needs in Temporal Annotation: a Proposal of Extension of ISO TimeML that Preserves Upward Compatibility. In LREC 2016, Portorož, Slov… find
  9. MERLOT (2016). Annotation scheme for the merlot french clinical corpus. Technical report. find
  10. Muzerelle, J., A. Lefeuvre, E. Schang, J.-Y. Antoine, A. Pelletier, D. Maurel, I. Eshkol, and J. Villaneau (2014). ANCOR Centre, a Large Free Spoken French Coreference Corpus: description of the Resource and Reliability Measures. In LREC 2014, Reyjavik, Iceland. find our card
  11. Pustejovsky, J., J. M. Castao, R. Ingria, R. Sauri, R. J. Gaizauskas, A. Setzer, G. Katz, and D. R. Radev (2003). Timeml: Robust specification of event and temporal expressions in text. In M. T. Maybury (Ed.), New Directions in Question Answering, pp. 28–34. AAAI Press. find
  12. Pustejovsky, J., J. Littman, and R. Saurı́ (2006). Argument structure in timeml. In Dagstuhl Seminar Proceedings. Schloss Dagstuhl-Leibniz-Zentrum für Informatik. find
  13. Reichenbach, H. (1947). Elements of Symbolic Logic. New York, Macmillan. find
  14. Zinsmeister, H. and S. Dipper (2011). Towards a Standard for Annotating Abstract Anaphora. find
  15. This licence is inherited from the original oral corpus. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/antoine-2017-temporal-odil-project-adapting-iso/