Elamur Research
☰
home / papers / 2017 / The Parallel Meaning Bank: Towards a Multilingual Corpus of Translations Annotated with Compositional Meaning Representations

The Parallel Meaning Bank: Towards a Multilingual Corpus of Translations Annotated with Compositional Meaning Representations

What it is about

The Parallel Meaning Bank (PMB) annotates English texts with compositional meaning representations based on Discourse Representation Theory, and projects the annotation to translations through word alignment. The assumption is that a translation keeps the meaning, so one annotation serves all its versions. The English annotation runs through segmentation, CCG parsing, semantic tags, symbolisation and lexical semantics, with manual correction at each layer. The corpus at the time held over 11 million tokens in 285,154 documents from twelve sources, among them subtitle-style short texts, news, Sherlock Holmes stories and the Bible. About 72% of documents have a German version, 14% Dutch and 42% Italian. The paper states three goals: a test bed for cross-lingual compositional semantics, data for parsers in other languages, and the detection of meaning shifts that translators make on purpose.

Ideas

For Mova

The code is tools/translate/meta.rs and the author delta. I1 and I3 address open problem 8 (style deltas) and open problem 6 (the standard leans to the Bibles). The paper's CCG and DRT pipeline is not used by Mova.

Library summary

The Research library's own summary of the paper, in our words — not the paper's text.

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/E17-2039.pdf

Summary

The authors aim to obtain formal semantic representations for several languages without costly manual annotation from scratch. To this end, they build the Parallel Meaning Bank (PMB), a corpus of parallel translations in English, German, Italian, and Dutch. The idea is as follows: if a translation preserves meaning, then the original and the translation should have the same logical representation. The English text is therefore annotated automatically, humans correct the errors, and the annotation is then projected onto the other languages via word alignment. The pipeline has five steps: tokenization, CCG-based syntactic parsing, universal semantic tagging, "symbolization" (lemmatization combined with normalization), and compositional DRS construction following Discourse Representation Theory. The models are the same for all languages and are retrained on manual corrections, which are called Bits of Wisdom. Depending on how much manual verification an annotation has undergone, it is assigned to the gold, silver, or bronze class. The authors consider the first results for Dutch promising. The value of the work is that the PMB is an open resource for cross-lingual compositional semantics, for training semantic parsers for languages other than English, and for detecting meaning shifts in translations.

Key points

  • The corpus contains over 11.3 million tokens in 285,154 documents from twelve sources, including Tatoeba, News-Commentary (OPUS), RTE, the Sherlock Holmes stories, and the Bible. An English version exists for all documents, a German one for 72%, an Italian one for 42%, and a Dutch one for 14%. Only 5% of documents are available in all four languages.
  • Tools: the Elephant tokenizer (character-level IOB tagging), the EasyCCG parser, whose grammar replaces type-changing rules with explicit empty elements, and Boxer, adapted to work with semtags instead of English POS tags. GIZA++ is used for word alignment.
  • The universal semantic tag set consists of 80 fine-grained tags grouped into 13 classes. A tagger based on deep residual networks achieved an accuracy of 83.6%.
  • The symbolizer is built as a character-level sequence-to-sequence model. It additionally draws on WordNet, Wikipedia, and the UNECE trade codes.
  • As of 13.02.2017, the gold class contained 6,810 English documents at the token level. The data are available via the web interface pmb.let.rug.nl.

Topics: Machine translation

Cite

The paper itself, not this card. BibTeX from the library's record of the work. arXiv 1702.03964

@inproceedings{abzianidze2017parallel,
  title = {{The Parallel Meaning Bank: Towards a Multilingual Corpus of Translations Annotated with Compositional Meaning Representations}},
  author = {Abzianidze, Lasha and Bjerva, Johannes and Evang, Kilian and Haagsma, Hessel and Noord, Rik van and Ludmann, Pierre and Nguyen, Duc-Duy and Bos, Johan},
  booktitle = {Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers},
  year = {2017},
  eprint = {1702.03964},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1702.03964}
}

References

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

  1. Kilian Evang. 2016. Cross-lingual Semantic Parsing with Categorial Grammars. Ph.D. thesis, University of Groningen. find
  2. Laura Banarescu, Claire Bonial, Shu Cai, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Kevin Knight, Philipp Koehn, Martha Palmer, and Nathan Schneider. 2013. Abstract Meaning Representation for sembanking. In Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse, pages 178–186, Sofia, Bulgaria. find our card
  3. Christiane Fellbaum, editor. 1998. WordNet. An Electronic Lexical Database. The MIT Press, Cambridge, Ma., USA. find
  4. Danilo Giampiccolo, Bernardo Magnini, Ido Dagan, and Bill Dolan. 2007. The third PASCAL Recognizing Textual Entailment challenge. In Proceedings of the ACL-PASCAL Workshop on Textual Entailment and Paraphrasing, pages 1–9. find
  5. Valerio Basile, Johan Bos, Kilian Evang, and Noortje Venhuizen. 2012. A platform for collaborative semantic annotation. In Proceedings of the Demonstrations at the 13th Conference of the European Chapter of the Association for Computational Linguistics (EACL 2012), pages 92–96, Avignon, France. find
  6. Hans Kamp and Uwe Reyle. 1993. From Discourse to Logic; An Introduction to Modeltheoretic Semantics of Natural Language, Formal Logic and DRT. Kluwer, Dordrecht. find
  7. Johannes Bjerva, Barbara Plank, and Johan Bos. 2016. Semantic tagging with deep residual networks. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3531–3541, Osaka, Japan. find
  8. Adam Kilgarriff. 1997. “I don’t believe in word senses”. Computers and the Humanities, 31(2):91– 113. find our card
  9. Arthur Langeveld. 1986. Vertalen wat er staat. Synthese, De Arbeiderspers. find
  10. Claire Bonial, William J. Corvey, Martha Palmer, Volha Petukhova, and Harry Bunt. 2011. A hierarchical unification of LIRICS and VerbNet semantic roles. In Proceedings of the 5th IEEE International Conference on Semantic Computing (ICSC 2011), pages 483–489. find
  11. Mike Lewis and Mark Steedman. 2014. A* CCG parsing with a supertag-factored model. In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 990–1000, Doha, Qatar. find our card
  12. Johan Bos, Stephen Clark, Mark Steedman, James R. Curran, and Julia Hockenmaier. 2004. Wide-coverage semantic representations from a CCG parser. In Proceedings of the 20th International Conference on Computational Linguistics (COLING 2004), pages 1240–1246, Geneva, Switzerland. find
  13. Mike Lewis, Luheng He, and Luke Zettlemoyer. 2015. Joint A* CCG parsing and semantic role labeling. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1444–1454. find
  14. Guido Minnen, John Carroll, and Darren Pearce. 2001. Applied morphological processing of English. Natural Language Engineering, 7(3):207–223. find
  15. Johan Bos, Valerio Basile, Kilian Evang, Noortje Venhuizen, and Johannes Bjerva. 2017. The Groningen Meaning Bank. In Nancy Ide and James Pustejovsky, editors, Handbook of Linguistic Annotation. Springer Netherlands. find
  16. Franz Josef Och and Hermann Ney. 2003. A systematic comparison of various statistical alignment models. Computational Linguistics, 29(1):19–51. find
  17. Johan Bos. 2015. Open-domain semantic parsing with Boxer. In Beáta Megyesi, editor, Proceedings of the 20th Nordic Conference of Computational Linguistics (NODALIDA 2015), pages 301–304. find
  18. Mark Steedman. 2001. The Syntactic Process. The MIT Press, Cambridge, Ma., USA. find
  19. Jörg Tiedemann. 2012. Parallel data, tools and interfaces in OPUS. In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC 2012), pages 2214–2218, Istanbul, Turkey. find
  20. Christos Christodouloupoulos and Mark Steedman. 2015. A massively parallel corpus: the Bible in 100 languages. Language Resources and Evaluation, 49(2):375–395. find our card
  21. Kilian Evang and Johan Bos. 2016. Cross-lingual learning of an open-domain semantic parser. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 579–588, Osaka, Japan. find
  22. Noortje Joost Venhuizen. 2015. Projection in Discourse: A data-driven formal semantic analysis. Ph.D. thesis, University of Groningen. find
  23. Nianwen Xue, Ondrej Bojar, Jan Hajic, Martha Palmer, Zdenka Uresova, and Xiuhong Zhang. 2014. Not an interlingua, but close: Comparison of English AMRs to Chinese and Czech. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC 2014), volume 14, pages 1765–1772. find our card
  24. Kilian Evang, Valerio Basile, Grzegorz Chrupała, and Johan Bos. 2013. Elephant: Sequence labeling for word and sentence segmentation. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1422– 1426, Seattle, Washington, USA. find