Elamur Research⌕
☰
home / papers / 2004 / BiFrameNet: Bilingual Frame Semantics Resource Construction by Cross-lingual Induction

BiFrameNet: Bilingual Frame Semantics Resource Construction by Cross-lingual Induction

Summary

Chinese had no frame-semantic resource comparable to the English FrameNet, which limited semantic transfer in English–Chinese machine translation. The authors build BiFrameNet automatically, without manual seed mappings, by linking English FrameNet lexical entries to Chinese word senses (concepts) in HowNet. The method has three stages:

  1. It proposes candidate links through two bilingual lexicons (HowNet and the LDC dictionary), keeping only pairs whose part of speech matches.
  2. It disambiguates by letting the other lexical entries of the same frame vote for HowNet categories.
  3. It recovers wrongly discarded senses using a HowNet taxonomy similarity measure.

The resource also needs Chinese example sentences. For these, the authors align Chinese and English POS sequences with dynamic programming, using POS transfer costs learned from a parallel corpus. Each aligned Chinese word inherits the semantic role of its English counterpart.

On highly ambiguous entries, lexical alignment reaches about 82% average F-measure. Role annotation of the induced Chinese sentences is roughly 68–72% accurate. The result is a bilingual semantic lexicon that can support statistical MT, transfer-based MT and computer-aided translation.

Key points

  • Resources: Berkeley FrameNet (over 7000 English lexical items with role-annotated examples), Chinese HowNet, the LDC English–Chinese dictionary, and the Hong Kong News parallel corpus.
  • Lexical mapping: Disambiguation keeps the top N=3 HowNet categories per frame. Pruned categories are restored through a taxonomy similarity score (Liu and Li, 2002) with α=1.6.
  • Lexical evaluation: Tested on the most ambiguous entries, each with 124–145 candidate HowNet links.
    • Large frames reach an overall F-measure of 87.4%, up from 52.3% after step 1 alone.
    • Small frames (24% of frames) reach 64.9%.
    • The weighted average is about 82%; beat.v alone scores 89.72.
  • Example sentences: POS tagging used the Ratnaparkhi maximum entropy tagger for English and an HMM tagger (ICTCLAS) for Chinese. In the cause_harm frame, role annotation accuracy is 68% on average and 71.8% for the top 100 sentences ranked by DP score.

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: Semantic role labeling · Part-of-speech tagging

Cited by

10 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 6 not yet.

Cite

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

@inproceedings{fung2004biframenet,
  title = {{BiFrameNet: Bilingual Frame Semantics Resource Construction by Cross-lingual Induction}},
  author = {Fung, Pascale and Chen, Benfeng},
  booktitle = {COLING 2004: Proceedings of the 20th International Conference on Computational Linguistics},
  year = {2004},
  url = {https://aclanthology.org/C04-1134/}
}

References

14 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Collin F. Baker, Charles J. Fillmore and John B. Lowe. (1998).The Berkeley FrameNet project. In Proceedings of the COLING-ACL, Montreal, Canada. find our card
  2. Hans C. Boas. Bilingual FrameNet Dictionaries for Machine Translation. In Proceedings of the Third International Conference on Language Resources and Evaluation. Las Palmas, Spain. Vol. IV: 1364-1371 2002. find our card
  3. Silviu Cucerzan and David Yarowsky. Bootstrapping a multilingual part-of-speech tagger in one person-day. In Proceedings of the Sixth Conference on Natural Language Learning (CoNLL). Taipei, Taiwan. 2002. find
  4. Dong, Zhendong., and Dong, Qiang. HowNet [online 2002]. Available at http://www.keenage.com/zhiwang/e_zhiwang.html ↗
  5. Bonnie J. Dorr, Gina-Anne Levow, and Dekang Lin.(2002).Construction of a Chinese-English Verb Lexicon for Machine Translation. In Machine Translation, Special Issue on Embedded MT, 17:1-2. find
  6. Michael Fleischman, Namhee Kwon and Eduard Hovy. Maximum Entropy Models for FrameNet Classification. In Proceedings of ACL 2003, Sapporo. find our card
  7. Daniel Gildea and Daniel Jurafsky.(2002).Automatic Labeling of Semantic Roles. In Computational Linguistics, Vol 28.3: 245-288. find our card
  8. Judy Illes and Wendy S. Francis. Convergent cortical representation of semantic processing in bilinguals. In Brain and Language, 70(3):347-363, 1999. find
  9. Liu Qun and Li, Sujian. Word Similarity Computing Based on How-net. In Computational Linguistics and Chinese Language Processing，Vol.7, No.2, August 2002, pp.59-76 find
  10. Grace Ngai, Marine Carpuat, and Pascale Fung. Identifying Concepts Across Languages: A First Step towards a Corpus-based Approach to Automatic Ontology Alignment. In Proceedings of COLING-02, Taipei, Taiwan. Martha Palmer and Wu Zhibiao. Verb Semantics for In Machine English-Chinese Translation. Tra… find
  11. Sabine Ploux and Hyungsuk Ji. A Model for Matching Semantic Maps between Languages (French/English, English/French). In Computational Linguistics 29(2):155-178, 2003. find
  12. Adwait Ratnaparkhi. A Maximum Entropy Part-Of-Speech Tagger. In Proceedings of EMNLP 2003, May 17-18, 1996. University of Pennsylvania Satoko Ikeda. Manual response set in a stroop-like task involving categorization of English and Japanese words indicates a common semantic representation. In Percept… find
  13. Dekai Wu. Stochastic inversion transduction grammars and bilingual parsing of parallel corpora. In Computational Linguistics 23(3):377-404, Sep 1997 find our card
  14. Dekai Wu. The HKUST leading question translation system. MT-Summit 2003. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/fung-2004-biframenet-bilingual-frame-semantics-resource/