Elamur Research⌕
☰
home / papers / 2003 / Japanese Zero Pronoun Resolution based on Ranking Rules and Machine Learning

Japanese Zero Pronoun Resolution based on Ranking Rules and Machine Learning

Summary

The paper addresses resolving Japanese zero pronouns, which are omitted arguments that give no number or gender cues, in full-text newspaper articles. Hand-written heuristics are hard to scale, and machine learning needs a lot of annotated data, so the authors combine the two. Candidate antecedents are first ordered lexicographically with simple ranking rules: semantic violation checked against Nihongo Goi Taikei, demotion of candidates inside relative clauses, Kameyama's property-sharing, distance, and Walker-style salience. An SVM then filters out unsuitable candidates, and the top-ranked surviving candidate is chosen. Experiments use Seki's annotated data from the Kyoto University Corpus 2.0, which the authors corrected, together with ChaSen and CaboCha output. The simple ordering alone outperformed the SVM alone, and the combination did better than either. Inspecting the feature weights showed that new features for complex sentences matter. The work shows that cheap linguistic preferences and a learned filter complement each other when training data is scarce.

Key points

  • Data: 30 general news articles (347 zeros) and 30 editorials (514 zeros). Editorials are harder; they average 28.7 sentences per article versus 13.9 for general news.
  • Preprocessing: ChaSen 2.2.9 and CaboCha 0.34 replaced the corpus's own annotations. Sentences were split at conjunctive postpositions, and only the ga, wo and ni cases were handled.
  • Semantic constraints: These come from Goi Taikei, which has 14,730 translation patterns for 6,103 verbs and about 300,000 words in roughly 3,000 categories.
  • Best configurations:
    • "vrads+svm2" performed best on general articles.
    • "davrs+svm2" performed best on editorials.
    • svm2 is a modified SVM for unbalanced data with a linear kernel and maxDi set to 3.
    • The ordinary SVM, maximum entropy and C4.5 all rejected too many candidates.
  • New features and cost:
    • Removing the new features Parallel, Unfinished and Intra lowered the editorial score by about 4 points.
    • Training on about 5,000–6,100 examples took 5–18 seconds.

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: Coreference resolution

Cited by

22 papers of our library cite it (of the 140673 whose bibliographies we read); 14 of them have a page here, 8 not yet.

Cite

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

@inproceedings{isozaki2003japanese,
  title = {{Japanese Zero Pronoun Resolution based on Ranking Rules and Machine Learning}},
  author = {Isozaki, Hideki and Hirao, Tsutomu},
  booktitle = {Proceedings of the 2003 Conference on Empirical Methods in Natural Language Processing},
  year = {2003},
  doi = {10.3115/1119355.1119379},
  url = {https://doi.org/10.3115/1119355.1119379}
}

References

23 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Chinatsu Aone and Scott William Bennett. 1995. Evaluating automated and manual acquisition of anaphora resolution strategies. In Proc. of ACL-1995, pages 122–129. find our card
  2. Terumasa Ehara and Yeun-Bae Kim. 1996. Zero-subject resolution by probabilistic model (in Japanese). Journal of Natural Language Processing, 3(4):67–86. find
  3. Barbara J. Grosz, Aravind K. Joshi, and Scott Weinstein. 1995. Centering: A framework for modelling the local coherence of discourse. Computational Linguistics, 21(2):203–226. find
  4. Tsutomu Hirao, Hideki Isozaki, Eisaku Maeda, and Yuji Matsumoto. 2002. Extracting important sentences with support vector machines. In Proc. of COLING-2002, pages 342–348. find
  5. Ryu Iida, Kentaro Inui, Hiroya Takamura, and Yuji Matsumoto. 2003a. Incorporating contextual cues in trainable models for coreference resolution (to appear). In Proc. of EACL Workshop on the Computational Treatment of Anaphora. find
  6. Ryu Iida, Kentaro Inui, Hiroya Takamura, and Yuji Matsumoto. 2003b. One method for resolving Japanese zero pronouns with machine learning model (in Japanese). In IPSJ SIG-NL 154. find
  7. Satoru Ikehara, Masahiro Miyazaki, Satoshi Shirai, Akio Yokoo, Hiromi Nakaiwa, Kentaro Ogura, Yoshifumi Ooyama, and Yoshihiko Hayashi. 1997. Goi-Taikei — A Japanese Lexicon (in Japanese). Iwanami Shoten. find
  8. Hideki Isozaki and Hideto Kazawa. 2002. Efficient support vector classifiers for named entity recognition. In Proc. of COLING-2002, pages 390–396. find
  9. Megumi Kameyama. 1986. A property-sharing constraint in centering. In Proc. of ACL-1986, pages 200– 206. find
  10. Taku Kudo and Yuji Matsumoto. 2001. Chunking with support vector machines. In Proc. of NAACL-2001, pages 192–199. find
  11. Katharina Morik, Peter Brockhausen, and Thorsten Joachims. 1999. Combining statistical learning with a knowledge-based approach — a case study in intensive care monitoring. In Proc. of ICML-1999, pages 268–277. find
  12. Masaki Murata and Makoto Nagao. 1997. An estimate of referents of pronouns in Japanese sentences using examples and surface expressions (in Japanese). Journal of Natural Language Processing, 4(1):41–56. find
  13. Hiromi Nakaiwa and Satoru Ikehara. 1993. Zero pronoun resolution in a Japanese to English machine translation system using verbal semantic attributes (in Japanese). Transaction of the Information Processing Society of Japan, 34(8):1705–1715. find
  14. Hiromi Nakaiwa and Satoru Ikehara. 1996. Intrasentential resolution of Japanese zero pronouns using pragmatic and semantic constraints (in Japanese). Journal of Natural Language Processing, 3(4):49–65. find
  15. Manabu Okumura and Kouji Tamura. 1996. Zero pronoun resolution based on centering theory. In Proc. of COLING-1996, pages 871–876. find
  16. Kazuhiro Seki, Atsushi Fujii, and Tetsuya Ishikawa. 2002a. Japanese zero pronoun resolution using a probabilistic model (in Japanese). Journal of Natural Language Processing, 9(3):63–85. find
  17. Kazuhiro Seki, Atsushi Fujii, and Tetsuya Ishikawa. 2002b. A probabilistic method for analyzing Japanese anaphora integrating zero pronoun detection and resolution. In Proc. of COLING-2002, pages 911–917. find our card
  18. Vladimir N. Vapnik. 1995. The Nature of Statistical Learning Theory. Springer. find
  19. Marilyn Walker, Masayo Iida, and Sharon Cote. 1994. Japanese discourse and the process of centering. Computational Linguistics, 20(2):193–233. find
  20. Kazuhide Yamamoto, Eiichiro Sumita, Osamu Furuse, and Hitoshi Iida. 1997. Ellipsis resolution in dialogues via decision-tree learning. In Proc. of NLPRS-1997, pages 423–428. find
  21. Mitsuko Yamura-Takei, Miho Fujiwara, Makoto Yoshie, and Teruaki Aizawa. 2002. Automatic linguistic analysis for language teachers: The case of zeros. In Proc. of COLING-2002, pages 1114–1120. find
  22. Kei Yoshimoto. 1986. Study of Japanese zero pronouns in discourse processing (in Japanese). In IPSJ SIG notes, NL-56-4, pages 1–8. find
  23. Keiichi Yoshino. 2001. Anaphora resolution of Japanese zero pronouns using machine learning (in Japanese). Master’s thesis, Nara Institute of Science and Technology. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/isozaki-2003-japanese-zero-pronoun-resolution-based/