Beyond Identity Coreference: Contrasting Indicators of Textual Coherence in English and German

What it is about

The paper treats coreference as chains that carry more than identity: type-instance, part-whole, hyponymy, repetition and situation anaphora ("That's difficult" referring to a clause). It measures four chain features in a parallel English-German corpus: chain number, chain length, the distance between elements, and the semantic relation between neighbouring elements. The corpus has written registers (political essays, popular-science articles, fiction) and spoken interviews, with about 121,000 English and 136,000 German tokens in total. The main finding is that register matters much more than language. Register explains about 60% of the variance in chain length, about 50–60% in chain number and 30–60% in distance (depending on chain type), while language explains 2% or less. Fiction has the longest coreference chains and the largest distances between elements. Essays have the most repetition, and popular-science texts have many hyponymy and meronymy relations. The authors argue that identity alone does not describe coherence in some registers.

Ideas

• I1. Compute the four chain features on Mova's own English texts from the coreference output, plus lexical chains from the lexicon's hyponym and meronym links in en/src/dict.rs. Report chain number, length, distance and relation type per text, as a deterministic table, with the register label from the text's metadata.
• I2. Use the chain distance as a topic-continuity score for the story world (world/src/story.rs): short distances between elements of one chain mean the topic is kept, long distances mean interwoven topics. Store the score with the chain IDs so the answer can show which chain it used.
• I3. Negative control: shuffle the sentences of a text and recompute the features. Distances and chain lengths should change, and the continuity score should fall; if they do not, the feature is not measuring order.
• I4. Abstract anaphor chains: link a demonstrative clause-level reference ("That is why") to the preceding clause, as the paper's example (2) does, and count how many chains this adds that identity-only coreference would miss.

For Mova

The chapters are 35. Reference and coreference (identity, bridging and situation anaphora), 36. Discourse relations and coherence (how chains support discourse relations) and 37. Register and variation (register as the main source of variation). The code is coref/src/sieve.rs and coref/src/mention.rs (the identity chains), world/src/story.rs (topic continuity over the story) and en/src/dict.rs (the lexical relations). The measures: chain features per register on Mova's English texts, run twice for identical output, the shuffled-order negative control, and the share of chains that identity-only coreference misses. The paper's corpus is not Mova data, so the numbers serve only as a comparison of pattern (register dominates language).

Topics: Coreference resolution

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/w16-0704

@inproceedings{kunz2016beyond,
  title = {{Beyond Identity Coreference: Contrasting Indicators of Textual Coherence in English and German}},
  author = {Kunz, Kerstin and Lapshinova-Koltunski, Ekaterina and Martínez, José Manuel},
  booktitle = {Proceedings of the Workshop on Coreference Resolution Beyond OntoNotes (CORBON 2016)},
  year = {2016},
  doi = {10.18653/v1/w16-0704},
  url = {https://doi.org/10.18653/v1/w16-0704}
}

References

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

1. M. Ariel. 2001. Accessibility theory: An overview. In T. Sanders, J. Schliperoord, and W. Spooren, editors, Text representation: Linguistic and psycholinguistic aspects, pages 29–87. John Benjamins Publishing, Amsterdam, Philadelphia. find
2. R. Harald Baayen. 2008. Analyzing Linguistic Data. A Practical Introduction to Statistics Using R. Cambridge University Press. find
3. John Chambers, William Cleveland, Beat Kleiner, and Paul Tukey. 1983. Graphical Methods for Data Analysis. Wadsworth. find
4. G. Doddington, A. Mitchell, M. Przybocki, L. Ramshaw, S. Strassel, and R. Weischedel. 2004. Automatic content extraction (ace) program – task definitions and performance measures. In Proceedings of LREC-2004: 4th International Conference on Language Resources and Evaluation, Lisbon, Portugal, May 24-30. find
5. Michael J. Greenacre. 2007. Correspondence analysis in practice. Chapman & Hall/CRC, Boca Raton. find
6. E. Hajičová, B. Hladká, and L. Kučová. 2006. An annotated corpus as a test bed for discourse structure analysis. In Proceedings of the Workshop on Constraints in Discourse, pages 82–89, Maynooth, Ireland. National University of Ireland. find
7. Silvia Hansen-Schirra, Stella Neumann, and Erich Steiner. 2012. Cross-linguistic Corpora for the Study of Translations. Insights from the Language Pair English-German. de Gruyter, Berlin, New York. find
8. A. A. Kibrik. 2011. Reference in discourse. Oxford University Press, Oxford. find
9. Kerstin Kunz and Ekaterina Lapshinova-Koltunski. 2015. Cross-linguistic analysis of discourse variation across registers. Special Issue of Nordic Journal of English Studies, 14(1):258–288. find
10. K. Lambrecht. 1994. Information structure and sentence form. Topic, focus and the mental representation of discourse referents. Cambridge University Press, Cambridge. find
11. Ekaterina Lapshinova-Koltunski and Kerstin Kunz. 2014a. Annotating cohesion for multillingual analysis. In Proceedings of the 10th Joint ACL - ISO Workshop on Interoperable Semantic Annotation, Reykjavik, Iceland, May. LREC. find
12. Ekaterina Lapshinova-Koltunski and Kerstin Kunz. 2014b. Annotating cohesion for multillingual analysis. In Proceedings of the 10th Joint ACL - ISO Workshop on Interoperable Semantic Annotation, Reykjavik, Iceland, May. LREC. find
13. Ekaterina Lapshinova-Koltunski, Kerstin Kunz, and Mar-ilisa Amoia. 2012. Compiling a multilingual spoken corpus. In Tommaso Raso Heliana Mello, Massimo Pettorino, editor, Proceedings of the VIIth GSCP International Conference: Speech and corpora, pages 79–84, Firenze. Firenze University Press. find
14. S. Pradhan, L. Ramshaw, M Mitchell, M. Palmer, R. Weischedel, and N. Xue. 2011. Conll-2011 shared task: Modeling unrestricted coreference in ontonotes. In Proceedings of the Fifteenth Conference on Computational Natural Language Learning: Shared Task. find our card
15. M. Strube and U. Hahn. 1999. Functional centering: Grounding referential coherence in information structure. Computational Linguistics, 25(3):309–344. find
16. BBN Technologies. 2006. Coreference Guidelines for English OntoNotes – Version 6.0. Linguistic Data Consortium. BBN Pronoun Coreference and Entity Type Corpus. find
17. H. Telljohann, E. Hinrichs, S. Kbler, H Zinsmeister, and K. Beck. 2012. Stylebook for the tübingen treebank of written german (tüba-d/z). Technical report, Tübingen, Germany. Revised Version. find
18. William N. Venables and David M. Smith. 2010. An Introduction to R. Notes on R: A Programming Environment for Data Analysis and Graphics. find
19. Š. Zikánová, E. Hajičová, B. Hladká, Jı́nová P., J. Mı́rovský, A. Nedoluzhko, L. Poláková, K. Rysová, M. Rysová, and J. Václ. 2015. Discourse and Coherence. From the Sentence Structure to Relations in Text. ÚFAL, Prague, Czech Republic. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/kunz-2016-beyond-identity-coreference-contrasting-indicators/