Elamur Research⌕
☰
home / papers / 2012 / Cross-Lingual Identification of Ambiguous Discourse Connectives for Resource-Poor Language

Cross-Lingual Identification of Ambiguous Discourse Connectives for Resource-Poor Language

Summary

The paper tackles the identification of explicit discourse connectives in Chinese, a language that lacked a public discourse-annotated corpus. It addresses two kinds of ambiguity: whether a word is used as a connective at all, and which relation sense it signals. The authors propose a language-independent, cross-lingual pipeline built on English resources. First, English connectives from PDTB2 are translated with a bilingual dictionary and expanded with Chinese synonyms. Next, each Chinese candidate's sense ambiguity is scored as the entropy of a relation distribution. That distribution is projected from the English connectives aligned to the candidate in parallel data, and a threshold (max-e) filters out the noisiest candidates. Finally, a co-training scheme with an English view and a Chinese view separates discourse from non-discourse uses. It starts from PDTB2 labels and gradually absorbs unlabeled Chinese instances. For evaluation, the authors built DTBC, a new Chinese discourse corpus. The filtered lexicon and the co-training classifier clearly beat dictionary-only and machine-translation baselines. This shows that useful discourse tools can be bootstrapped for resource-poor languages.

Key points

  • DTBC: 500 news articles from Penn Chinese Tree Bank 6 (CTB6), annotated with 2,549 explicit relations, including connectives and arguments.
  • Lexicon quality: the entropy-filtered lexicon (DIC+ENT, 231 entries) reached F = 0.7773 (precision 0.8615). The raw dictionary list (561 entries) scored F = 0.6675. Over 30% of Chinese candidates turned out to be sense-ambiguous.
  • Explicit relation classification on DTBC: DIC+ENT achieved F = 0.7481 with max-e = 0.3, tuned on 20% of DTBC. The best baseline, DIC-1, reached F = 0.6827, and the MT-based approach reached F = 0.5825.
  • Discourse-usage identification: co-training with added lexical and word-alignment features (COT-2) obtained accuracy 0.7428 and F = 0.7933. The connective-only baseline scored 0.2470 / 0.3961, and MT-2 scored 0.6404 / 0.6814. Frequent words such as 和 and 在 rarely function as connectives, which makes this task harder in Chinese than in English.
  • Tools and resources: the pipeline uses PDTB2 top-level senses (Temporal, Contingency, Comparison, Expansion), Baidu Translator, Stanford Parser, MALLET, and the extended Chinese synonym list of Che et al. (2010).

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: Discourse analysis

Cited by

9 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 7 not yet.

Cite

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

@inproceedings{zhou2012cross,
  title = {{Cross-Lingual Identification of Ambiguous Discourse Connectives for Resource-Poor Language}},
  author = {Zhou, Lanjun and Gao, Wei and Li, Binyang and Wei, Zhongyu and Wong, Kam-Fai},
  booktitle = {Proceedings of COLING 2012: Posters},
  year = {2012},
  url = {https://aclanthology.org/C12-2138/}
}

References

19 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Al-Saif, A. and Markert, K. (2010). The leeds arabic discourse treebank: Annotating discourse connectives for arabic. In Language Resources and Evaluation Conference (LREC). find our card
  2. Alsaif, A. and Markert, K. (2011). Modelling discourse relations for arabic. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 736-- 747. find our card
  3. Carlson, L., Marcu, D., and Okurowski, M. (2001). Building a discourse-tagged corpus in the framework of rhetorical structure theory. In Proceedings of the Second SIGdial Workshop on Discourse and Dialogue-Volume 16, pages 1--10. Association for Computational Linguistics. find our card
  4. Che, W., Li, Z., and Liu, T. (2010). Ltp: A chinese language technology platform. In Proceedings of the 23rd International Conference on Computational Linguistics: Demonstrations, pages 13--16. Association for Computational Linguistics. da Cunha, I., Torres-Moreno, J., and Sierra, G. (2011). On the… find
  5. Mann, W. and Thompson, S. (1988). Rhetorical structure theory: Toward a functional theory of text organization. Text-Interdisciplinary Journal for the Study of Discourse, 8(3):243--281. find our card
  6. McKeown, K. (1992). Text generation: using discourse strategies and focus constraints to generate natural language text. Cambridge Univ Pr. find
  7. Miltsakaki, E., Prasad, R., Joshi, A., and Webber, B. (2004). The penn discourse treebank. In Proceedings of the 4th International Conference on Language Resources and Evaluation. Citeseer. find our card
  8. Och, F. and Ney, H. (2003). A systematic comparison of various statistical alignment models. Computational linguistics, 29(1):19--51. find
  9. Pitler, E. and Nenkova, A. (2009a). Using syntax to disambiguate explicit discourse connectives in text. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 13--16, Suntec, Singapore. Association for Computational Linguistics. find our card
  10. Pitler, E. and Nenkova, A. (2009b). Using syntax to disambiguate explicit discourse connectives in text. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 13--16. Association for Computational Linguistics. find our card
  11. Pitler, E., Raghupathy, M., Mehta, H., Nenkova, A., Lee, A., and Joshi, A. (2008). Easily identifiable discourse relations. Proceedings of COLING 2008, Posters Proceedings, pages 87--90. find our card
  12. Prasad, R., Dinesh, N., Lee, A., Miltsakaki, E., Robaldo, L., Joshi, A., and Webber, B. (2008a). The penn discourse treebank 2.0. In Proceedings of the 6th International Conference on Language Resources and Evaluation (LREC 2008), pages 2961--2968. Citeseer. find our card
  13. Prasad, R., Husain, S., Sharma, D., and Joshi, A. (2008b). Towards an annotated corpus of discourse relations in hindi. Proceedings of IJCNLP-2008. find our card
  14. Spärck Jones, K. (2007). Automatic summarising: The state of the art. Information Processing & Management, 43(6):1449--1481. find
  15. Versley, Y. (2010). Discovery of ambiguous and unambiguous discourse connectives via annotation projection. In Proceedings of Workshop on Annotation and Exploitation of Parallel Corpora (AEPC), pages 83--82. find
  16. Xue, N. (2005). Annotating discourse connectives in the chinese treebank. In Proceedings of the Workshop on Frontiers in Corpus Annotations II: Pie in the Sky, pages 84--91. Association for Computational Linguistics. find our card
  17. Xue, N., Xia, F., Chiou, F., and Palmer, M. (2005). The penn chinese treebank: Phrase structure annotation of a large corpus. Natural Language Engineering, 11(02):207--238. find
  18. Zhou, L., Li, B., Gao, W., Wei, Z., and Wong, K. (2011). Unsupervised discovery of discourse relations for eliminating intra-sentence polarity ambiguities. In Proceedings of the 2011 Conference on Empirical methods in natural language processing, pages 162--171. Association for Computational Linguis… find our card
  19. Zhou, Y. and Xue, N. (2012). Pdtb-style discourse annotation of chinese text. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 69--77, Jeju Island, Korea. Association for Computational Linguistics. find

↑ back to top

This page in Markdown →