An Information Bottleneck Approach for Controlling Conciseness in Rationale Extraction

By Bhargavi Paranjape, Mandar Joshi, John Thickstun, Hannaneh Hajishirzi, Luke S. Zettlemoyer. 2020

115 citations recorded by the source corpus.

Decisions of complex language understanding models can be rationalized by limiting their inputs to a relevant subsequence of the original text. A rationale should be as concise as possible without significantly degrading task performance, but this balance can be difficult to achieve in practice. In this paper, we show that it is possible to better manage this trade-off by optimizing a bound on the Information Bottleneck (IB) objective. Our fully unsupervised approach jointly learns an explainer that predicts sparse binary masks over sentences, and an end-task predictor that considers only the extracted rationale. Using IB, we derive a learning objective that allows direct control of mask sparsity levels through a tunable sparse prior. Experiments on ERASER benchmark tasks demonstrate significant gains over norm-minimization techniques for both task performance and agreement with human rationales. Furthermore, we find that in the semi-supervised setting, a modest amount of gold rationales (25% of training examples) closes the gap with a model that uses the full input.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Few-Shot Self-Rationalization with Natural Language Prompts (2021, methodology): “Since then, many highlighting methods have been designed following their approach (Bastings et al., 2019; Yu et al., 2019; Paranjape et al., 2020).”
Honest Students from Untrusted Teachers: Learning an Interpretable Question-Answering Pipeline from a Pretrained Language Model (2022, background): “In contrast, prior work has found that previous unsupervised techniques for constructing rationales (Paranjape et al., 2020; Guerreiro and Martins, 2021) decreased performance by 10-20 F 1 on SQuAD (Chen et al., 2022).”
Evaluating and Characterizing Human Rationales (2020, methodology): “Popular techniques include LIME (Ribeiro et al., 2016), integrated gradients (Sundararajan et al., 2017), SHAP (Lundberg and Lee, 2017), and attention mechanisms (Lei et al., 2016; Paranjape et al., 2020).”
Cross-Domain Transfer of Generative Explanations Using Text-to-Text Models (2021, background): “Another line of work for explainable NLP is based on rationalization pipelines that aim to produce extractive explanations by splitting the prediction process into two subsequent modules; a rationale extractor and a predictor [1,10,25,34].”
Challenges and Strategies in Cross-Cultural NLP (2022, background): “ple of the former is beer reviews, a go-to domain in sentiment analysis (Zeng et al., 2019; Ji et al., 2020; Paranjape et al., 2020).”
Right for the Right Reason: Evidence Extraction for Trustworthy Tabular Reasoning (2022, background): “However, these models ignore the causal relationship between evidence selection and label prediction (Herzig et al., 2021; Paranjape et al., 2020).”