Fine-grained Sentiment Analysis with Faithful Attention

By Ruiqi Zhong, Steven Shao, K. McKeown. 2019

53 citations recorded by the source corpus.

While the general task of textual sentiment classification has been widely studied, much less research looks specifically at sentiment between a specified source and target. To tackle this problem, we experimented with a state-of-the-art relation extraction model. Surprisingly, we found that despite reasonable performance, the model's attention was often systematically misaligned with the words that contribute to sentiment. Thus, we directly trained the model's attention with human rationales and improved our model performance by a robust 4~8 points on all tasks we defined on our data sets. We also present a rigorous analysis of the model's attention, both trained and untrained, using novel and intuitive metrics. Our results show that untrained attention does not provide faithful explanations; however, trained attention with concisely annotated human rationales not only increases performance, but also brings faithful explanations. Encouragingly, a small amount of annotated human rationales suffice to correct the attention in our task.

Subfields: Artificial Intelligence

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

Citations in the corpus

BERTology Meets Biology: Interpreting Attention in Protein Language Models (2020, background): “Interpreting attention on natural language sequences is a well-established area of research [12, 30, 87, 90].”
ERASER: A Benchmark to Evaluate Rationalized NLP Models (2019, background): “Attention fares better in terms of AUPRC — suggesting better agreement with human rationales — which is also in line with prior ﬁndings that it may provide plausible, but not faithful, explanation (Zhong et al., 2019).”
Evaluating Explanations: How Much Do Explanations from the Teacher Aid Students? (2020, methodology): “Several past approaches have used attention regularization to incorporate human rationales, with an aim to improve the overall performance of the system for classification tasks (Bao et al., 2018; Zhong et al., 2019) and machine translation (Yin et al.”
Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates (2021, background): “The models can achieve better performance or learn more efficiently if supervisions are provided for these feature importance scores (Ross et al., 2017; Zhong et al., 2019; Pruthi et al., 2020a).”
Explaining and Improving Model Behavior with k Nearest Neighbor Representations (2020, background): “Zhong, Shao, and McKeown (2019) train attention to directly attend to human rationales and obtain more faithful explanations.”
Learning to Faithfully Rationalize by Construction (2020, methodology): “NLP for constructing ‘heatmaps’ over , provide robust, (Feng et al., Wallace, 2019; Wiegreffe and Pinter, 2019; Ser-rano and Smith, 2019; Brunner et al., 2020; Zhong et al., 2019; Pruthi et al., 2020).”