Elamur Research
☰
home / papers / 2020 / Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data

Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/2020.acl-main.463.pdf

Summary

Large pretrained language models such as BERT and GPT-2 do well on many tasks that seem to require understanding. Both research papers and the popular press often say these models "understand" language or capture meaning. In this ACL 2020 position paper, Bender and Koller argue that these claims mix up linguistic form with meaning. They define form as any observable trace of language: text, pixels, bytes or articulator movements. They define meaning as the link between an expression and something outside language, namely the speaker's communicative intent or the expression's conventional meaning. Building on these definitions, they argue on theoretical grounds that a system trained only to predict strings gets no signal for learning that link. Their support comes from thought experiments, research on how children acquire language, and a discussion of grounding in distributional semantics. The paper presents no new datasets or models. Its contribution is methodological: researchers should use words like "understanding" precisely, design tasks and report results with the form–meaning distinction in mind, and not treat benchmark gains as near-solutions to human-analogous NLU.

Key points

  • Formal definitions: meaning is the relation M ⊆ E × I, pairing expressions with communicative intents. Conventional meaning is C ⊆ E × S, pairing expressions with standing meanings that can be interpreted against a world. "Understanding" means recovering i from e.
  • BERTology evidence: probing studies such as Tenney et al. (2019) and Hewitt and Manning (2019) on ELMo and BERT show that LMs encode formal structure like agreement and dependency structure. Other results point the other way:
    • Niven and Kao (2019) showed that BERT's results on the Argument Reasoning Comprehension Task drop to chance once adversarial negated examples are added.
    • McCoy et al. (2019) found that BERT's performance on MNLI depends on lexical-overlap heuristics and falls well below chance on data built to defeat them.
  • Thought experiments: an octopus (O) eavesdrops on two people, A and B, and learns their message patterns but cannot respond sensibly to new situations. A model trained on all Java code on GitHub could not learn what programs actually do. Unit tests would give it partial grounding, but only if its designers built in a way to interpret them.
  • Answers to counterarguments: more data does not solve the problem, because new communicative intents keep appearing. Neural representations have no interpretations. If unsupervised MT trained only on monolingual corpora (Lample et al., 2018) matched supervised MT, the authors say this would mean accurate translation does not require understanding.
  • Conclusion: LMs learn a useful but partial reflection of meaning in form. The authors call for precise wording, humility and a top-down view so the field picks the right hill to climb toward NLU.

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

339 papers of our library cite it (of the 140673 whose bibliographies we read); 18 of them have a page here, 321 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2020.acl-main.463

@inproceedings{bender2020climbing,
  title = {{Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data}},
  author = {Bender, Emily M. and Koller, Alexander},
  booktitle = {Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
  year = {2020},
  doi = {10.18653/v1/2020.acl-main.463},
  url = {https://doi.org/10.18653/v1/2020.acl-main.463}
}

References

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

  1. deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4171–4186, Minneapolis, Minnesota. Association for Co… find
  2. Yossi Adi, Einat Kermany, Yonatan Belinkov, Ofer Lavi, and Yoav Goldberg. 2017. Fine-grained analysis of sentence embeddings using auxiliary prediction tasks. In Proceedings of ICLR. find our card
  3. Dare A. Baldwin. 1995. Understanding the link between joint attention and language. In Chris Moore and Philip J. Dunham, editors, Joint Attention: Its Origins and Role in Development, pages 131–158. Psychology Press. find
  4. Andrei Barbu, David Mayo, Julian Alverio, William Luo, Christopher Wang, Dan Gutfreund, Josh Tenenbaum, and Boris Katz. 2019. ObjectNet: A large-scale bias-controlled dataset for pushing the limits of object recognition models. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d’Alché Buc, E. Fox, a… find
  5. Marco Baroni, Raffaella Bernardi, Roberto Zamparelli, et al. 2014. Frege in space: A program for compositional distributional semantics. Linguistic Issues in Language Technology, 9(6):5–110. find
  6. Yonatan Bisk, Ari Holtzman, Jesse Thomason, Jacob Andreas, Yoshua Bengio, Joyce Chai, Mirella Lapata, Angeliki Lazaridou, Jonathan May, Aleksandr Nisnevich, Nicolas Pinto, and Joseph Turian. 2020. Experience grounds language. ArXiv preprint. find our card
  7. Ned Block. 1981. Psychologism and behaviorism. The Philosophical Review, 90(1):5–43. find
  8. Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. 2015. A large annotated corpus for learning natural language inference. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 632–642, Lisbon, Portugal. Association for Computati… find
  9. Rechele Brooks and Andrew N. Meltzoff. 2005. The development of gaze following and its relation to language. Developmental Science, 8(6):535–543. find
  10. Herbert H. Clark. 1996. Using Language. Cambridge University Press, Cambridge. find
  11. Stephen Clark. 2015. Vector space models of lexical meaning. In Shalom Lappin and Chris Fox, editors, Handbook of Contemporary Semantic Theory, second edition, pages 493–522. Wiley-Blackwell. find
  12. Dheeru Dua, Yizhong Wang, Pradeep Dasigi, Gabriel Stanovsky, Sameer Singh, and Matt Gardner. 2019. DROP: A reading comprehension benchmark requiring discrete reasoning over paragraphs. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistic… find our card
  13. Guy Emerson. 2020. What are the goals of distributional semantics? In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, Seattle, Washington. Association for Computational Linguistics. find
  14. Katrin Erk. 2016. What do you know about an alligator when you know the company it keeps? Semantics & Pragmatics, 9(17):1–63. find
  15. Allyson Ettinger, Ahmed Elgohary, Colin Phillips, and Philip Resnik. 2018. Assessing composition in sentence vector representations. In Proceedings of the 27th International Conference on Computational Linguistics, pages 1790–1801, Santa Fe, New Mexico, USA. Association for Computational Linguistics… find
  16. Yoav Goldberg. 2019. Assessing BERT’s syntactic abilities. ArXiv preprint. find our card
  17. H. Paul Grice. 1968. Utterer’s meaning, sentence-meaning, and word-meaning. Foundations of Language, 4(3):225–242. find
  18. Ivan Habernal, Henning Wachsmuth, Iryna Gurevych, and Benno Stein. 2018. The argument reasoning comprehension task: Identification and reconstruction of implicit warrants. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Lang… find
  19. Ido Dagan, Oren Glickman, and Bernardo Magnini. 2006. The PASCAL Recognising Textual Entailment Challenge. In Machine Learning Challenges. Evaluating Predictive Uncertainty, Visual Object Classification, and Recognising Tectual Entailment, pages 177–190, Berlin, Heidelberg. Springer. find
  20. William L. Hamilton, Jure Leskovec, and Dan Jurafsky. 2016. Diachronic word embeddings reveal statistical laws of semantic change. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1489–1501, Berlin, Germany. Association for Com… find our card
  21. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Stevan Harnad. 1990. The symbol grounding problem. Physica D, 42:335–346. find
  22. Junxian He, Graham Neubig, and Taylor Berg-Kirkpatrick. 2018. Unsupervised learning of syntactic structure with invertible neural projections. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 1292–1302, Brussels, Belgium. Association for Computational… find
  23. Benjamin Heinzerling. 2019. NLP’s Clever Hans moment has arrived. Blog post, accessed 12/4/2019. find
  24. Aurelie Herbelot. 2013. What is in a text, what isn’t, and what this has to do with lexical semantics. In Proceedings of the 10th International Conference on Computational Semantics (IWCS 2013) – Short Papers, pages 321–327, Potsdam, Germany. Association for Computational Linguistics. find
  25. Aurélie Herbelot, Eva von Redecker, and Johanna Müller. 2012. Distributional techniques for philosophical enquiry. In Proceedings of the 6th Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities, pages 45–54, Avignon, France. Association for Computational Linguisti… find
  26. John Hewitt and Christopher D. Manning. 2019. A structural probe for finding syntax in word representations. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 4129… find our card
  27. MD. Zakir Hossain, Ferdous Sohel, Mohd Fairuz Shi-ratuddin, and Hamid Laga. 2019. A comprehensive survey of deep learning for image captioning. ACM Comput. Surv., 51(6):118:1–118:36. find
  28. Ganesh Jawahar, Benoı̂t Sagot, and Djamé Seddah. 2019. What does BERT learn about the structure of language? In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3651–3657, Florence, Italy. Association for Computational Linguistics. find our card
  29. Nitish Shirish Keskar, Bryan McCann, Lav R. Varshney, Caiming Xiong, and Richard Socher. 2019. CTRL: A conditional transformer language model for controllable generation. ArXiv preprint. find our card
  30. Douwe Kiela, Luana Bulat, and Stephen Clark. 2015. Grounding semantics in olfactory perception. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 231– 23… find
  31. Douwe Kiela and Stephen Clark. 2015. Multi- and cross-modal semantics beyond vision: Grounding in auditory perception. In Proceedings of the 2015 find
  32. Conference on Empirical Methods in Natural Language Processing, pages 2461–2470, Lisbon, Portugal. Association for Computational Linguistics. find
  33. Alexander Koller, Konstantina Garoufi, Maria Staudte, and Matthew Crocker. 2012. Enhancing referential success by tracking hearer gaze. In Proceedings of the 13th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 30–39, Seoul, South Korea. Association for Computational Li… find
  34. Patricia K. Kuhl. 2007. by the social brain? 10(1):110–120. find
  35. Is speech learning ‘gated’ Developmental Science, Guillaume Lample, Myle Ott, Alexis Conneau, Ludovic Denoyer, and Marc’Aurelio Ranzato. 2018. Phrase-based & neural unsupervised machine translation. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 5039… find
  36. Chu-Cheng Lin, Waleed Ammar, Chris Dyer, and Lori Levin. 2015. Unsupervised POS induction with word embeddings. In Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1311–1316, Denver, Colorado. Associ… find
  37. Sally McConnell-Ginet. 1984. The origins of sexist language in discourse. Annals of the New York Academy of Sciences, 433(1):123–135. find
  38. Tom McCoy, Ellie Pavlick, and Tal Linzen. 2019. Right for the wrong reasons: Diagnosing syntactic heuristics in natural language inference. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3428–3448, Florence, Italy. Association for Computational Ling… find our card
  39. Daniel McDuff and Ashish Kapoor. 2019. Visceral machines: Reinforcement learning with intrinsic physiological rewards. In International Conference on Learning Representations. find
  40. Tomas Mikolov, Wen-tau Yih, and Geoffrey Zweig. 2013. Linguistic regularities in continuous space word representations. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 746–751, Atlanta, Georgia.… find
  41. Timothy Niven and Hung-Yu Kao. 2019. Probing neural network comprehension of natural language arguments. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4658–4664, Florence, Italy. Association for Computational Linguistics. find
  42. Stephan Oepen, Marco Kuhlmann, Yusuke Miyao, Daniel Zeman, Silvie Cinková, Dan Flickinger, Jan Hajič, and Zdeňka Urešová. 2015. find
  43. SemEval 2015 Task 18: Broad-coverage semantic dependency parsing. In Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015). find
  44. Yasuhito Ohsugi, Itsumi Saito, Kyosuke Nishida, Hisako Asano, and Junji Tomita. 2019. A simple but effective method to incorporate multi-turn context with BERT for conversational machine comprehension. In Proceedings of the First Workshop on NLP for Conversational AI, pages 11–17, Florence, Italy. A… find
  45. Simon Ostermann, Michael Roth, and Manfred Pinkal. 2019. MCScript2.0: A machine comprehension corpus focused on script events and participants. In Proceedings of the Eighth Joint Conference on Lexical and Computational Semantics (*SEM 2019), pages 103–117, Minneapolis, Minnesota. Association for Com… find
  46. Fabio Petroni, Tim Rocktäschel, Sebastian Riedel, Patrick Lewis, Anton Bakhtin, Yuxiang Wu, and Alexander Miller. 2019. Language models as knowledge bases? In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural… find our card
  47. W. V. O. Quine. 1960. Word and Object. MIT Press. find
  48. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. Open AI Blog, accessed 12/4/2019. find
  49. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100,000+ questions for machine comprehension of text. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2383–2392, Austin, Texas. Association for Computational Linguistics. find our card
  50. Michael J. Reddy. 1979. The conduit metaphor: A case of frame conflict in our language about language. In A. Ortony, editor, Metaphor and Thought, pages 284–310. Cambridge University Press. find
  51. Herbert Rubenstein and John B. Goodenough. 1965. Contextual correlates of synonymy. Communications of the ACM, 8(10):627–633. find
  52. John Searle. 1980. Minds, brains, and programs. Behavioral and Brain Sciences, 3(3):417–457. find
  53. Catherine E Snow, Anjo Arlman-Rupp, Yvonne Hass-ing, Jan Jobse, Jan Joosten, and Jan Vorster. 1976. Mothers’ speech in three social classes. Journal of Psycholinguistic Research, 5(1):1–20. find
  54. Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Sam Bowman, Dipanjan Das, and Ellie Pavlick. 2019. What do you learn from context? Probing for sentence structure in contextualized word representations. In International Conference on Lear… find our card
  55. Michael Tomasello and Michael Jeffrey Farrar. 1986. Joint attention and early language. Child Development, 57(6):1454–1463. find
  56. Alan Turing. 1950. Computing machinery and intelligence. Mind, 59(236):433–460. find
  57. Alex Wang, Yada Pruksachatkun, Nikita Nangia, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel Bowman. 2019. SuperGLUE: A stickier benchmark for general-purpose language understanding systems. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d’Alché Buc, E. Fox, and R. Garnett, ed… find our card
  58. Alex Warstadt, Yu Cao, Ioana Grosu, Wei Peng, Hagen Blix, Yining Nie, Anna Alsop, Shikha Bordia, Haokun Liu, Alicia Parrish, Sheng-Fu Wang, Jason Phang, Anhad Mohananey, Phu Mon Htut, Paloma Jeretic, and Samuel R. Bowman. 2019. Investigating BERT’s knowledge of language: Five analysis methods with N… find
  59. Joseph Weizenbaum. 1966. ELIZA—A computer program for the study of natural language communication between men and machines. Communications of the ACM, 9:36–45. find
  60. Jason Weston, Antoine Bordes, Sumit Chopra, Alexander M. Rush, Bart van Merriënboer, Armand Joulin, and Tomas Mikolov. 2016. Towards AI-complete question answering: A set of prerequisite toy tasks. In Proceedings of ICLR. find our card
  61. Adina Williams, Nikita Nangia, and Samuel Bowman. 2018. A broad-coverage challenge corpus for sentence understanding through inference. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Pa… find
  62. Ludwig Wittgenstein. 1953. Philosophical Investigations. MacMillan, New York. find
  63. Thomas Wolf. 2018. Learning meaning in natural language processing — The semantics mega-thread. Blog post, accessed 4/15/2020. find