CoChat

Understanding Commonsense Inference Aptitude of Deep Contextual Representations

Open access
Da, Jeff, Kasai, Jungo · 2019 · Proceedings of the First Workshop on Commonsense Inference in Natural Language Processing

The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.

Abstract

Pretrained deep contextual representations have advanced the state-of-the-art on various commonsense NLP tasks, but we lack a concrete understanding of the capability of these models.Thus, we investigate and challenge several aspects of BERT's commonsense representation abilities.First, we probe BERT's ability to classify various object attributes, demonstrating that BERT shows a strong ability in encoding various commonsense features in its embedding space, but is still deficient in many areas.Next, we show that, by augmenting BERT's pretraining data with additional data related to the deficient attributes, we are able to improve performance on a downstream commonsense reasoning task while using a minimal amount of data.Finally, we develop a method of fine-tuning knowledge graphs embeddings alongside BERT and show the continued importance of explicit knowledge graphs.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.18653/v1/d19-6001

Cites

35 of 35 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Swarm plots showing attribute fit scores for GloVe (left) and BERT (right). Each dot represents a single attribute, displayed along the x-axis according to the classifier's ability to fit that feature with the given embeddings. The y-axis is not significant, and instead, dots are displaced along the y-axis instead of overlapping to show quantity. The median fit score per embedding type is displayed with a dotted line.
Figure 1 Swarm plots showing attribute fit scores for GloVe (left) and BERT (right). Each dot represents a single attribute, displayed along the x-axis according to the classifier's ability to fit that feature with the given embeddings. The y-axis is not significant, and instead, dots are displaced along the y-axis instead of overlapping to show quantity. The median fit score per embedding type is displayed with a dotted line.CC-BY · From the PDF
Figure 2 Differences between fit scores when using GloVe (start of arrow) or BERT (end of arrows) embeddings.
Figure 2 Differences between fit scores when using GloVe (start of arrow) or BERT (end of arrows) embeddings.CC-BY · From the PDF
Figure 3 Linear regression fit of accuracy on MCScript 2.0, per attribute, versus fit score, with the inner 90 percent bootstrap confidence intervals highlighted (n = 1000). Each dot represents the accuracy of questions related to one attribute.
Figure 3 Linear regression fit of accuracy on MCScript 2.0, per attribute, versus fit score, with the inner 90 percent bootstrap confidence intervals highlighted (n = 1000). Each dot represents the accuracy of questions related to one attribute.CC-BY · From the PDF
Figure 4 Outline of our baseline method of improving BERT for commonsense reasoning. Our method fine-tunes BERT through multiple facets while optimizing for accuracy and reduced train steps. We use RACE (Lai et al., 2017) as an external dataset, and MCScript 2.0 (Ostermann et al., 2019) as our downstream task.
Figure 4 Outline of our baseline method of improving BERT for commonsense reasoning. Our method fine-tunes BERT through multiple facets while optimizing for accuracy and reduced train steps. We use RACE (Lai et al., 2017) as an external dataset, and MCScript 2.0 (Ostermann et al., 2019) as our downstream task.CC-BY · From the PDF
Figure 5 Visualization of ConceptNet knowledge base queries. The word eat is being queried with the other words in the text, with the valid edges discovered displayed against the left.
Figure 5 Visualization of ConceptNet knowledge base queries. The word eat is being queried with the other words in the text, with the valid edges discovered displayed against the left.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Research works printed but not linked: 5

  1. Elia Bruni, Gemma Boleda, Marco Baroni, and Nam- Khanh Tran. 2012. Distributional semantics in tech- nicolor. In Proc. of ACL.
  2. Matthew Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettlemoyer. 2018. Deep contextualized word rep- resentations. In Proc. of NAACL-HLT.
  3. Melissa Roemmele, Cosmin Adrian Bejan, and An- drew S. Gordon. 2011. Choice of plausible alter- natives: An evaluation of commonsense causal rea- soning. In AAAI 2011 Spring Symposium.
  4. Lenhart Schubert. 2015. What kinds of knowledge are needed for genuine understanding? In Proc of Cog- nitum.
  5. Zhengyan Zhang, Xu Han, Zhiyuan Liu, Xin Jiang, Maosong Sun, and Qun Liu. 2019. ERNIE: En- hanced language representation with informative en- tities. In Proc. of ACL, Florence, Italy. Association for Computational Linguistics.

Cited by

See all 40

Metadata

Authors
Da, J., & Kasai, J.
Affiliations
University of Washington (US)
Venue
Proceedings of the First Workshop on Commonsense Inference in Natural Language Processing
Published
2019
Pages
1-12
OpenAlex
W2985797697
Cited by
22
References
34
Citations by year
2024: 1 · 2023: 2 · 2022: 2 · 2021: 13 · 2020: 4
FWCI
1.82
Type
conference paper
Language
en
Open access
Licence
cc-by
Funders
National Science Foundation
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
Keywords
commonsense reasoning, BERT, pre-trained language models, context representation, knowledge graph embedding, fine-tuning, embedding space analysis

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.