CoChat

Rationalizing Neural Predictions

Open access
Lei, Tao, Barzilay, Regina, Jaakkola, Tommi · 2016 · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing

The DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index.

Abstract

Prediction without justification has limited applicability.As a remedy, we learn to extract pieces of input text as justifications -rationales -that are tailored to be short and coherent, yet sufficient for making the same prediction.Our approach combines two modular components, generator and encoder, which are trained to operate well together.The generator specifies a distribution over text fragments as candidate rationales and these are passed through the encoder for prediction.Rationales are never given during training.Instead, the model is regularized by desiderata for rationales.We evaluate the approach on multi-aspect sentiment analysis against manually annotated test cases.Our approach outperforms attention-based baseline by a significant margin.We also successfully illustrate the method on the question retrieval task. 1

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/d16-1011

Cites

33 of 33 references verified0 citation errors · 0 not checkable

Statistics

Figures

Image 1
CC-BY · From the PDF
Figure 1 An example of a review with ranking in two categories. The rationale for Look prediction is shown in bold.
Figure 1 An example of a review with ranking in two categories. The rationale for Look prediction is shown in bold.CC-BY · From the PDF
Figure 2 Mean squared error of all aspects on the test set (yaxis) when various percentages of text are extracted as rationales (x-axis). 220k training data is used.
Figure 2 Mean squared error of all aspects on the test set (yaxis) when various percentages of text are extracted as rationales (x-axis). 220k training data is used.CC-BY · From the PDF
Figure 4 Precision (y-axis) when various percentages of text are extracted as rationales (x-axis) for the appearance aspect.
Figure 4 Precision (y-axis) when various percentages of text are extracted as rationales (x-axis) for the appearance aspect.CC-BY · From the PDF
Figure 5 Learning curves of the optimized cost function on the development set and the precision of rationales on the test set. The smell (aroma) aspect is the target aspect.
Figure 5 Learning curves of the optimized cost function on the development set and the precision of rationales on the test set. The smell (aroma) aspect is the target aspect.CC-BY · From the PDF
Figure 6 Retrieval MAP on the test set when various percentages of the texts are chosen as rationales. Data points correspond to models trained with different hyper-parameters.
Figure 6 Retrieval MAP on the test set when various percentages of the texts are chosen as rationales. Data points correspond to models trained with different hyper-parameters.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: 7

  1. [6]6. Cicero dos Santos, Luciano Barbosa, Dasha Bogdanova, and Bianca Zadrozny. 2015. Learning hybrid rep- resentations to retrieve semantically equivalent ques- tions. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Lan- guage Processing (Volume 2: Short Papers), pages 694-699, Beijing, China, July. Association for Com- putational Linguistics.
  2. [7]7. Manaal Faruqui, Jesse Dodge, Sujay K. Jauhar, Chris Dyer, Eduard Hovy, and Noah A. Smith. 2015a. Retrofitting word vectors to semantic lexicons. In Pro- ceedings of NAACL.
  3. [9]9. Aurélie Herbelot and Eva Maria Vecchi. 2015. Build- ing a shared world: mapping distributional to model- theoretic semantic spaces. In Proceedings of the 2015 Conference on Empirical Methods in Natural Lan- guage Processing. Association for Computational Lin- guistics.
  4. [12]12. Mohit Iyyer, Varun Manjunatha, Jordan Boyd-Graber, and Hal Daumé III. 2015. Deep unordered compo- sition rivals syntactic methods for text classification. In Proceedings of the 53rd Annual Meeting of the As- sociation for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers).
  5. [17]17. Yoon Kim. 2014. Convolutional neural networks for sen- tence classification. In Proceedings of the Empiricial Methods in Natural Language Processing (EMNLP 2014).
  6. [27]27. Tim Rocktäschel, Edward Grefenstette, Karl Moritz Her- mann, Tomáš Kočiskỳ, and Phil Blunsom. 2016. Rea- soning about entailment with neural attention. In In- ternational Conference on Learning Representations.
  7. [29]29. Sebastian Thrun. 1995. Extracting rules from artifi- cial neural networks with distributed representations. In Advances in neural information processing systems (NIPS).

Cited by

See all 900

Metadata

Authors
Lei, T., Barzilay, R., & Jaakkola, T.
Affiliations
MIT Computer Science and Artificial Intelligence Laboratory (US)
Venue
Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing
Published
2016
Pages
107-117
OpenAlex
W2440159969
Cited by
777
References
27
Citations by year
2026: 27 · 2025: 33 · 2024: 39 · 2023: 88 · 2022: 92 · 2021: 138 · 2020: 126 · 2019: 101 · 2018: 78 · 2017: 39 · 2016: 4
FWCI
39.87
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
  • Conference on Empirical Methods in Natural Language Processing (EMNLP) (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion)
  • DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc-sa)
  • arXiv (Cornell University)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
  • Physical Sciences › Computer Science › Artificial Intelligence › Sentiment Analysis and Opinion Mining
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
Keywords
explanation generation, question retrieval, model regularization

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.