Elamur Research⌕
☰
home / papers / 2021 / Does External Knowledge Help Explainable Natural Language Inference? Automatic Evaluation vs. Human Ratings

Does External Knowledge Help Explainable Natural Language Inference? Automatic Evaluation vs. Human Ratings

Paper · 2021 · Hendrik Schuff, Hsiu-Yu Yang, Heike Adel et al.

Summary

This paper asks whether external knowledge, which is known to raise NLI label accuracy, also improves the free-text explanations that explainable NLI systems produce. A second question is how a language model's implicit knowledge compares with symbolic resources. The base system (VANILLA) joins the ESIM encoder with the PRED-EXPL predictor of Camburu et al., using BERT embeddings fine-tuned on SNLI. The authors then build variants that inject knowledge from COMET, ConceptNet Numberbatch embeddings and GPT-2, in some cases combining several sources. All models are trained on e-SNLI and also tested on NLI stress-test sets that probe specific reasoning types.

External knowledge can raise both accuracy and BLEU, and fine-tuned language models score best overall. However, they fail on negation and numerical reasoning, and adding other sources can reduce that weakness. A large crowdsourcing study tells a different story from the automatic metrics: even big gaps in accuracy, BLEU or BLEURT do not produce significant differences in human ratings. This casts doubt on judging explainable systems by automatic scores alone.

Key points

  • Knowledge sources compared:

    • COMET, used as a natural-language knowledge base queried per ConceptNet relation;
    • ConceptNet Numberbatch embeddings;
    • GPT-2.

  • In total, 11 model architectures were evaluated on e-SNLI and NLI stress tests.

  • The largest score gaps between models were:

    • up to 3.23 points of label accuracy;
    • 10.17 BLEU;
    • 0.31 BLEURT.

    None of these gaps led to a significantly different human rating on any dimension.

  • The human study rated four dimensions: label, explanation, commonsense and grammar correctness.

    • FILTERED-ENS was rated significantly better on explanation correctness than VANILLA and COMET+CONT.
    • The e-SNLI ground-truth explanations ranked in the lower half on all four dimensions, suggesting e-SNLI is too noisy to separate strong models (WT5-11B was also included for comparison).

  • Model predictions and crowdsourced ratings are released at github.com/boschresearch/external-knowledge-explainable-nli.

Interesting ideas

  • I1. Instead of one overall score, humans judged an explanation on several separate axes: label, reasoning, use of commonsense and grammar. Each axis can then fail independently.
  • I2. The gold references were put into the human study as an ordinary condition, which revealed that the reference data itself rates poorly.
  • I3. Before querying COMET, sentences are split into noun and verb sub-phrases using POS patterns. The generated knowledge phrases are then filtered by Sentence-BERT similarity to the source text.
  • I4. One alignment rule forces two words to be linked when they are both related in a knowledge base and already attended to by the model, so symbolic knowledge confirms soft evidence.
  • I5. Stress tests split by reasoning type showed that implicit language-model knowledge breaks on numbers and negation, while symbolic sources help cover those gaps.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (learn): Mova's explanation traces could be scored per axis, for example correct frame, correct world inference and fluent wording, so that a regression on one axis is not hidden by the others.
  • I2 (learn): Mova could add gold or silver answers as blind entries in its evaluations, a cheap way to detect noisy targets before training on them.
  • I3 (world): when the knowledge graph is queried, Mova could look up UD subtree chunks rather than whole sentences, and keep only retrieved facts that stay close to the text's entities.
  • I4 (answer): when Mova matches entities between a question and a story, a knowledge-graph relation could act as a required second witness before a match is accepted.
  • I5 (answer): Mova's school tests and maths word problems could report results per phenomenon, such as negation and quantities, to check that symbolic modules are covering the cases where LLM teachers are weak.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of StuttgartStuttgart · Germany75 works · 46,113 citations🏢Robert BoschStuttgart · Germany7 works · 2,234 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 2Research subjects and languages the work is about.Reasoning · 510 worksEvaluation & benchmarks · 685 works

Cite

Cite as Hendrik Schuff, Hsiu-Yu Yang, Heike Adel and Ngoc Thang Vu. 2021. Does External Knowledge Help Explainable Natural Language Inference? Automatic Evaluation vs. Human Ratings. Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2021.blackboxnlp-1.3

@inproceedings{schuff2021does,
  title = {{Does External Knowledge Help Explainable Natural Language Inference? Automatic Evaluation vs. Human Ratings}},
  author = {Hendrik Schuff, Hsiu-Yu Yang, Heike Adel and Ngoc Thang Vu},
  booktitle = {Proceedings of the Fourth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP},
  year = {2021},
  doi = {10.18653/v1/2021.blackboxnlp-1.3},
  url = {https://doi.org/10.18653/v1/2021.blackboxnlp-1.3}
}

References

37 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. pages 1657–1668, Vancouver, Canada. Association for Computational Linguistics. find
  2. Pepa Atanasova, Jakob Grue Simonsen, Christina Lioma, and Isabelle Augenstein. 2020. Generating fact checking explanations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7352–7364, Online. Association for Computational Linguistics. ↗
  3. Miruna-Adriana Clinciu, Arash Eshghi, and Helen Hastie. 2021. A study of automatic metrics for the evaluation of natural language explanations. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 2376–2387, Online. Association for Computational Linguistics. find
  4. Lisa Bauer, Lingjia Deng, and Mohit Bansal. 2021. ERNIE-NLI: Analyzing the Impact of Domain-Specific External Knowledge on Enhanced Representations for NLI. In Proceedings of Deep Learning Inside Out (DeeLIO): The 2nd Workshop on Knowledge Extraction and Integration for Deep Learning Architectures, pages 58–69, Online. Association for Computational Linguistics. find
  5. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of 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 Computational Linguistics. ↗
  6. Chandra Bhagavatula, Ronan Le Bras, Chaitanya Malaviya, Keisuke Sakaguchi, Ari Holtzman, Hannah Rashkin, Doug Downey, Wen-tau Yih, and Yejin Choi. 2020. Abductive Commonsense Reasoning. In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net. ↗
  7. Keyur Faldu, Amit P. Sheth, Prashant Kikani, and Hemang Akabari. 2021. KI-BERT: infusing knowledge context for better language and domain understanding. CoRR, abs/2104.08145. ↗
  8. Antoine Bosselut, Hannah Rashkin, Maarten Sap, Chaitanya Malaviya, Asli Celikyilmaz, and Yejin Choi. 2019. COMET: Commonsense transformers for automatic knowledge graph construction. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4762–4779, Florence, Italy. Association for Computational Linguistics.
  9. John Fox. 2003. Effect Displays in R for Generalised Linear Models. Journal of Statistical Software, 8(15). find
  10. Valentin Jijkoun and Maarten de Rijke. 2005. Recognizing textual entailment using lexical similarity. In Proceedings of the PASCAL Challenges Workshop on Recognising Textual Entailment, pages 73–76. find
  11. 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 Computational Linguistics. ↗
  12. Sawan Kumar and Partha Talukdar. 2020. NILE : Natural language inference with faithful natural language explanations. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 8730–8742, Online. Association for Computational Linguistics.
  13. Tao Lei, Regina Barzilay, and Tommi Jaakkola. 2016. Rationalizing neural predictions. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 107–117, Austin, Texas. Association for Computational Linguistics.
  14. Oana-Maria Camburu, Tim Rocktäschel, Thomas Lukasiewicz, and Phil Blunsom. 2018. e-snli: Natural language inference with natural language explanations. In Advances in Neural Information Processing Systems 31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018, December 3-8, 2018, Montréal, Canada, pages 9560– 9572. ↗
  15. Alexander Hanbo Li and Abhinav Sethy. 2019. Knowledge enhanced attention for robust natural language inference. CoRR, abs/1909.00102. ↗
  16. Qian Chen, Xiaodan Zhu, Zhen-Hua Ling, Diana Inkpen, and Si Wei. 2018. Neural natural language inference models enhanced with external knowledge. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2406–2417, Melbourne, Australia. Association for Computational Linguistics.
  17. Tao Li and Vivek Srikumar. 2019. Augmenting neural networks with first-order logic. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 292–302, Florence, Italy. Association for Computational Linguistics. ↗
  18. Qian Chen, Xiaodan Zhu, Zhen-Hua Ling, Si Wei, Hui Jiang, and Diana Inkpen. 2017. Enhanced LSTM for natural language inference. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Tianda Li, Xiaodan Zhu, Quan Liu, Qian Chen, Zhigang Chen, and Si Wei. 2019. Several experiments on investigating pretraining and knowledge-enhanced models for natural language inference. CoRR, abs/1904.12104. ↗
  19. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. OpenAI blog, (8):9. find
  20. Bill Yuchen Lin, Xinyue Chen, Jamin Chen, and Xiang Ren. 2019. KagNet: Knowledge-Aware Graph Networks for Commonsense Reasoning. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 2829–2839, Hong Kong, China. Association for Computational Linguistics. ↗
  21. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21:140:1–140:67. ↗
  22. Scott Lundberg and Su-In Lee. 2017. A unified approach to interpreting model predictions. volume abs/1705.07874. ↗
  23. Nazneen Fatema Rajani, Bryan McCann, Caiming Xiong, and Richard Socher. 2019. Explain yourself! leveraging language models for commonsense reasoning. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4932–4942, Florence, Italy. Association for Computational Linguistics.
  24. Bill MacCartney, Michel Galley, and Christopher D. Manning. 2008. A phrase-based alignment model for natural language inference. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 802–811, Honolulu, Hawaii. Association for Computational Linguistics. find
  25. Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3982–3992, Hong Kong, China. Association for Computational Linguistics. ↗
  26. Aakanksha Naik, Abhilasha Ravichander, Norman Sadeh, Carolyn Rose, and Graham Neubig. 2018. Stress test evaluation for natural language inference. In Proceedings of the 27th International Conference on Computational Linguistics, pages 2340–2353, Santa Fe, New Mexico, USA. Association for Computational Linguistics. ↗
  27. Marco Túlio Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. "why should I trust you?": Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, August 13-17, 2016, pages 1135–1144. ACM.
  28. Sharan Narang, Colin Raffel, Katherine Lee, Adam Roberts, Noah Fiedel, and Karishma Malkan. 2020. Wt5?! training text-to-text models to explain their predictions. CoRR, abs/2004.14546. ↗
  29. Xiaoman Pan, Kai Sun, Dian Yu, Jianshu Chen, Heng Ji, Claire Cardie, and Dong Yu. 2019. Improving question answering with external knowledge. In Proceedings of the 2nd Workshop on Machine Reading for Question Answering, MRQA@EMNLP 2019, Hong Kong, China, November 4, 2019, pages 27–37. Association for Computational Linguistics. ↗
  30. Hendrik Schuff, Heike Adel, and Ngoc Thang Vu. 2020. F1 is Not Enough! Models and Evaluation Towards User-Centered Explainable Question Answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7076–7095, Online. Association for Computational Linguistics. find
  31. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318, Philadelphia, Pennsylvania, USA. Association for Computational Linguistics. ↗
  32. Thibault Sellam, Dipanjan Das, and Ankur Parikh. 2020. BLEURT: Learning robust metrics for text generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7881–7892, Online. Association for Computational Linguistics. ↗
  33. 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 Language Processing (EMNLP-IJCNLP), pages 2463–2473, Hong Kong, China. Association for Computational Linguistics. ↗
  34. Dominic Seyler, Tatiana Dembelova, Luciano Del Corro, Johannes Hoffart, and Gerhard Weikum. 2018. A study of the importance of external knowledge in the named entity recognition task. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 241–246, Melbourne, Australia. Association for Computational Linguistics. find
  35. Matt Post. 2018. A call for clarity in reporting BLEU scores. In Proceedings of the Third Conference on Machine Translation: Research Papers, pages 186– 191, Brussels, Belgium. Association for Computational Linguistics. ↗
  36. Chen Shi, Shujie Liu, Shuo Ren, Shi Feng, Mu Li, Ming Zhou, Xu Sun, and Houfeng Wang. 2016. Knowledge-based semantic embedding for machine translation. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics B (Volume 1: Long Papers), pages 2245–2254, Berlin, Germany. Association for Computational Linguistics. Figure 4 shows an example of the study interface used to collect human ratings as discussed in Section 5. ↗
  37. Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2014. Deep inside convolutional networks: Visualising image classification models and saliency maps. In 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Workshop Track Proceedings. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/schuff-2021-does-external-knowledge-help-explainable/