Elamur Research⌕
☰
home / papers / 2014 / Metaphor Detection with Cross-Lingual Model Transfer

Metaphor Detection with Cross-Lingual Model Transfer

Summary

The paper tackles automatic detection of metaphor, deciding whether a short syntactic construction is meant literally or figuratively. It focuses on two construction types, subject-verb-object (SVO) and adjective-noun (AN). Corpus studies cited in the paper link these to most metaphoric expressions. The authors build an English classifier from coarse, concept-level features: abstractness and imageability scores, WordNet supersenses (plus an adjective supersense taxonomy), and unsupervised vector-space word representations. A random forest makes the decisions. To test whether metaphor is conceptual rather than tied to particular words, they translate Spanish, Farsi and Russian word tuples into English with bilingual dictionaries and apply the English model unchanged. They also release new metaphor-annotated datasets, including Russian and English corpora. The English system reaches state-of-the-art results, and the transferred model works reasonably well in all three other languages. This matters because it offers a route to metaphor detection for languages with few resources. It also supports the view that metaphors are conceptual mappings rather than lexical facts.

Key points

  • Inputs and features: each SVO triple or AN pair becomes a feature vector. The vector combines per-word conceptual features with pairwise conjunction features, built as the vectorized outer product of two words' feature vectors.
  • Abstractness and imageability scores: these are extended from the MRC psycholinguistic database (4,295 abstractness ratings and 1,156 imageability ratings) to the whole vocabulary.
    • Logistic regression over word vectors does the extension, with held-out accuracy of 0.94 (abstractness) and 0.85 (imageability).
    • The scores are binarized at thresholds of 0.8 and 0.9.
  • Supersenses: WordNet supplies 26 noun and 15 verb supersenses, and a 13-class adjective supersense taxonomy comes from Tsvetkov et al. (2014).
  • Embeddings and classifier: the word vectors were trained on the WMT-2011 news commentary corpus (180,834 types). Classification uses a random forest in scikit-learn, which slightly beat logistic regression and SVM.
  • Cross-lingual results (English-trained model):
    • For AN, ROC AUC was 0.92 for EN, 0.83 for FA, 0.80 for RU and 0.73 for ES.
    • F-scores with optimal thresholds fall roughly between 0.72 and 0.85 across languages and both construction types.
    • Code is on GitHub (ytsvetko/metaphor).

Interesting ideas

  • I1. Language-agnostic features for model transfer: by describing words with concepts instead of lexical identity, a model trained once in a resource-rich language can be reused elsewhere through a bilingual dictionary alone. This is a cheap route to a first Ukrainian baseline for semantic tasks lacking annotated data.
  • I2. Growing a small lexicon with embeddings: a few thousand human ratings can be spread across the full vocabulary by training a classifier on word vectors. This trick suits any sparse psycholinguistic or sentiment lexicon.
  • I3. Outer-product conjunction features: these capture interactions such as an "abstract subject + concrete verb" mismatch, which signals a selectional-preference violation. The same idea fits other relation-classification tasks where the combination matters more than either word alone.
  • I4. Random forests for small data: they need no hyperparameter tuning and tend not to overfit as trees are added. That makes them a sensible default when labeled data are scarce.
  • I5. Transfer as a hypothesis test: if an English-only model succeeds on other languages, that is evidence the phenomenon is conceptual rather than lexical. This evaluation design can probe other linguistic claims empirically.

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: Natural language processing

Cited by

142 papers of our library cite it (of the 140884 whose bibliographies we read); 7 of them have a page here, 135 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/v1/p14-1024

@inproceedings{tsvetkov2014metaphor,
  title = {{Metaphor Detection with Cross-Lingual Model Transfer}},
  author = {Tsvetkov, Yulia and Boytsov, Leonid and Gershman, Anatole and Nyberg, Eric and Dyer, Chris},
  booktitle = {Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2014},
  doi = {10.3115/v1/p14-1024},
  url = {https://doi.org/10.3115/v1/p14-1024}
}

References

43 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Marco Baroni, Raffaella Bernardi, Ngoc-Quynh Do, and Chung-chieh Shan. 2012. Entailment above the word level in distributional semantics. In Proc. of EACL, pages 23–32. find our card
  2. Julia Birke and Anoop Sarkar. 2007. Active learning for the identification of nonliteral language. In Proc. of the Workshop on Computational Approaches to Figurative Language, FigLanguages ’07, pages 21– 28. find
  3. Leo Breiman. 2001. Random forests. Machine Learning, 45(1):5–32. find
  4. George Aaron Broadwell, Umit Boz, Ignacio Cases, Tomek Strzalkowski, Laurie Feldman, Sarah Taylor, Samira Shaikh, Ting Liu, Kit Cho, and Nick Webb. 2013. Using imageability and topic chaining to locate metaphors in linguistic corpora. In Social Computing, Behavioral-Cultural Modeling and Prediction,… find
  5. Massimiliano Ciaramita and Yasemin Altun. 2006. Broad-coverage sense disambiguation and information extraction with a supersense sequence tagger. In Proc. of EMNLP, pages 594–602. find our card
  6. Scott C. Deerwester, Susan T Dumais, Thomas K. Landauer, George W. Furnas, and Richard A. Harshman. 1990. Indexing by latent semantic analysis. JASIS, 41(6):391–407. find
  7. Manaal Faruqui and Chris Dyer. 2014. Improving vector space word representations using multilingual correlation. In Proc. of EACL. Association for Computational Linguistics. find our card
  8. Christiane Fellbaum, editor. 1998. WordNet: An Electronic Lexical Database. Language, Speech and Communication. MIT Press. find
  9. Lisa Gandy, Nadji Allan, Mark Atallah, Ophir Frieder, Newton Howard, Sergey Kanareykin, Moshe Koppel, Mark Last, Yair Neuman, and Shlomo Argamon. 2013. Automatic identification of conceptual metaphors with limited knowledge. In Proc. of the Twenty-Seventh AAAI Conference on Artificial Intelligence,… find
  10. Matt Gedigian, John Bryant, Srini Narayanan, and Branimir Ciric. 2006. Catching metaphors. In Proceedings of the 3rd Workshop on Scalable Natural Language Understanding, pages 41–48. find
  11. Birgit Hamp and Helmut Feldweg. 1997. Germanet-a lexical-semantic net for German. In Proc. of ACL workshop Automatic Information Extraction and Building of Lexical Semantic Resources for NLP Applications, pages 9–15. find our card
  12. Dirk Hovy, Shashank Srivastava, Sujay Kumar Jauhar, Mrinmaya Sachan, Kartik Goyal, Huiying Li, Whitney Sanders, and Eduard Hovy. 2013. Identifying metaphorical word use with tree kernels. In Proc. of the First Workshop on Metaphor in NLP, page 52. find
  13. Eric H. Huang, Richard Socher, Christopher D. Manning, and Andrew Y. Ng. 2012. Improving word representations via global context and multiple word prototypes. In Proc. of ACL, pages 873–882. find our card
  14. Franz Hundsnurscher and Jochen Splett. 1982. Semantik der Adjektive des Deutschen. Number 3137. Westdeutscher Verlag. find
  15. Mikhail Kozhenikov and Ivan Titov. 2013. Cross-lingual transfer of semantic role labeling models. In Proc. of ACL, pages 1190–1200. find
  16. Saisuresh Krishnakumaran and Xiaojin Zhu. 2007. Hunting elusive metaphors using lexical resources. In Proc. of the Workshop on Computational approaches to Figurative Language, pages 13–20. find
  17. George Lakoff and Mark Johnson. 1980. Conceptual metaphor in everyday language. The Journal of Philosophy, pages 453–486. find
  18. Lori Levin, Teruko Mitamura, Davida Fromm, Brian MacWhinney, Jaime Carbonell, Weston Feely, Robert Frederking, Anatole Gershman, and Carlos Ramirez. 2014. Resources for the detection of conventionalized metaphors in four languages. In Proc. of LREC. find
  19. André F. T. Martins, Noah A. Smith, Eric P. Xing, Pedro M. Q. Aguiar, and Mário A. T. Figueiredo. 2010. Turbo parsers: dependency parsing by approximate variational inference. In Proc. of ENMLP, pages 34– 44. find our card
  20. Zachary J Mason. 2004. CorMet: a computational, corpus-based conventional metaphor extraction system. Computational Linguistics, 30(1):23–44. find
  21. Ryan McDonald, Slav Petrov, and Keith Hall. 2011. Multi-source transfer of delexicalized dependency parsers. In Proc. of EMNLP. find our card
  22. Tomas Mikolov, Quoc V. Le, and Ilya Sutskever. 2013. Exploiting similarities among languages for Machine Translation. CoRR, abs/1309.4168. find
  23. Roberto Navigli. 2009. Word sense disambiguation: A survey. ACM Comput. Surv., 41(2):10:1–10:69, February. find
  24. Yair Neuman, Dan Assaf, Yohai Cohen, Mark Last, Shlomo Argamon, Newton Howard, and Ophir Frieder. 2013. Metaphor identification in large texts corpora. PloS one, 8(4):e62343. find
  25. F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. 2011. Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:282… find
  26. Claudia Perlich, Foster Provost, and Jeffrey S. Simonoff. 2003. Tree induction vs. logistic regression: a learning-curve analysis. Journal of Machine Learning Research, 4:211–255. find
  27. Nathan Schneider, Behrang Mohit, Chris Dyer, Kemal Oflazer, and Noah A Smith. 2013. Supersense tagging for Arabic: the MT-in-the-middle attack. In Proc. of NAACL-HLT, pages 661–667. find
  28. Ekaterina Shutova and Lin Sun. 2013. Unsupervised metaphor identification using hierarchical graph factorization clustering. In Proc. of NAACL-HLT, pages 978–988. find
  29. Ekaterina Shutova and Simone Teufel. Metaphor corpus annotated for source-target domain mappings. In Proc. of LREC, pages 3255–3261. find our card
  30. Ekaterina Shutova, Lin Sun, and Anna Korhonen. 2010. Metaphor identification using verb and noun clustering. In Proc. of COLING, pages 1002–1010. find our card
  31. Ekaterina Shutova, Simone Teufel, and Anna Korhonen. 2013. Statistical metaphor processing. Computational Linguistics, 39(2):301–353. find
  32. Ekaterina Shutova. 2010. Models of metaphor in NLP. In Proc. of ACL, pages 688–697. find
  33. Gerard J Steen, Aletta G Dorst, J Berenike Herrmann, Anna A Kaal, and Tina Krennmayr. 2010. Metaphor in usage. Cognitive Linguistics, 21(4):765–796. find
  34. Tomek Strzalkowski, George Aaron Broadwell, Sarah Taylor, Laurie Feldman, Boris Yamrom, Samira Shaikh, Ting Liu, Kit Cho, Umit Boz, Ignacio Cases, et al. 2013. Robust extraction of metaphors from novel data. In Proc. of the First Workshop on Metaphor in NLP, page 67. find
  35. Oscar Täckström, Dipanjan Das, Slav Petrov, Ryan McDonald, and Joakim Nivre. 2013. Token and type constraints for cross-lingual part-of-speech tagging. TACL, 1:1–12. find
  36. Paul H Thibodeau and Lera Boroditsky. 2011. Metaphors we think with: The role of metaphor in reasoning. PLoS One, 6(2):e16782. find
  37. Yulia Tsvetkov, Elena Mukomel, and Anatole Gershman. 2013. Cross-lingual metaphor detection using common semantic features. In The 1st Workshop on Metaphor in NLP 2013, page 45. find
  38. Yulia Tsvetkov, Nathan Schneider, Dirk Hovy, Archna Bhatia, Manaal Faruqui, and Chris Dyer. 2014. Augmenting English adjective senses with supersenses. In Proc. of LREC. find
  39. Joseph Turian, Lev Ratinov, Yoshua Bengio, and Dan Roth. 2009. A preliminary evaluation of word representations for named-entity recognition. In NIPS Workshop on Grammar Induction, Representation of Language and Language Learning, pages 1–8. find
  40. Joseph Turian, Lev Ratinov, and Yoshua Bengio. 2010. Word representations: a simple and general method for semi-supervised learning. In Proc. of ACL, pages 384–394. find our card
  41. Peter D. Turney, Yair Neuman, Dan Assaf, and Yohai Cohen. 2011. Literal and metaphorical sense identification through concrete and abstract context. In Proc. of EMNL, pages 680–690. find our card
  42. Yorick Wilks. 1978. Making preferences more active. Artificial Intelligence, 11(3):197–223. 2010. find
  43. Michael Wilson. 1988. MRC Psycholinguistic Database: Machine-usable dictionary, version 2.00. Behavior Research Methods, Instruments, & Computers, 20(1):6–10. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tsvetkov-2014-metaphor-detection-cross-lingual-model/