We propose a new, more actionable view of neural network interpretability and data analysis by leveraging the remarkable matching effectiveness of representations derived from deep networks, guided by an approach for class-conditional feature detection. The decomposition of the filter-n-gram interactions of a convolutional neural network (CNN) and a linear layer over a pre-trained deep network yields a strong binary sequence labeler, with flexibility in producing predictions at—and defining loss functions for—varying label granularities, from the fully supervised sequence labeling setting to the challenging zero-shot sequence labeling setting, in which we seek token-level predictions but only have document-level labels for training. From this sequence-labeling layer we derive dense representations of the input that can then be matched to instances from training, or a support set with known labels. Such introspection with inference-time decision rules provides a means, in some settings, of making local updates to the model by altering the labels or instances in the support set without re-training the full model. Finally, we construct a particular K-nearest neighbors (K-NN) model from matched exemplar representations that approximates the original model’s predictions and is at least as effective a predictor with respect to the ground-truth labels. This additionally yields interpretable heuristics at the token level for determining when predictions are less likely to be reliable, and for screening input dissimilar to the support set. In effect, we show that we can transform the deep network into a simple weighting over exemplars and associated labels, yielding an introspectable—and modestly updatable—version of the original model.

In this position paper, we propose that the community consider encouraging researchers to include two riders, a “Lay Summary” and an “AI Safety Disclosure”, as part of future NLP papers published in ACL forums that present user-facing systems. The goal is to encourage researchers–via a relatively non-intrusive mechanism–to consider the societal implications of technologies carrying (un)known and/or (un)knowable long-term risks, to highlight failure cases, and to provide a mechanism by which the general public (and scientists in other disciplines) can more readily engage in the discussion in an informed manner. This simple proposal requires minimal additional up-front costs for researchers; the lay summary, at least, has significant precedence in the medical literature and other areas of science; and the proposal is aimed to supplement, rather than replace, existing approaches for encouraging researchers to consider the ethical implications of their work, such as those of the Collaborative Institutional Training Initiative (CITI) Program and institutional review boards (IRBs).

In a controlled experiment of sequence-to-sequence approaches for the task of sentence correction, we find that character-based models are generally more effective than word-based models and models that encode subword information via convolutions, and that modeling the output data as a series of diffs improves effectiveness over standard approaches. Our strongest sequence-to-sequence model improves over our strongest phrase-based statistical machine translation model, with access to the same data, by 6 M2 (0.5 GLEU) points. Additionally, in the data environment of the standard CoNLL-2014 setup, we demonstrate that modeling (and tuning against) diffs yields similar or better M2 scores with simpler models and/or significantly less data than previous sequence-to-sequence approaches.