Dissecting Generation Modes for Abstractive Summarization Models via Ablation and Attribution

TL;DR

This paper introduces a two-step interpretability framework for neural abstractive summarization models, combining ablation to identify generation modes with attribution methods to analyze decision sources, revealing memorization and complex phenomena.

Contribution

It presents a novel approach to dissect and interpret summarization models by categorizing generation modes and applying attribution techniques to understand decision origins.

Findings

Identifies when models rely on input versus language modeling.

Detects memorized phrases and their training origins.

Analyzes complex phenomena like sentence fusion at the instance level.

Abstract

Despite the prominence of neural abstractive summarization models, we know little about how they actually form summaries and how to understand where their decisions come from. We propose a two-step method to interpret summarization model decisions. We first analyze the model's behavior by ablating the full model to categorize each decoder decision into one of several generation modes: roughly, is the model behaving like a language model, is it relying heavily on the input, or is it somewhere in between? After isolating decisions that do depend on the input, we explore interpreting these decisions using several different attribution methods. We compare these techniques based on their ability to select content and reconstruct the model's predicted token from perturbations of the input, thus revealing whether highlighted attributions are truly important for the generation of the next token. While this machinery can be broadly useful even beyond summarization, we specifically demonstrate its capability to identify phrases the summarization model has memorized and determine where in the training pipeline this memorization happened, as well as study complex generation phenomena like sentence fusion on a per-instance basis.