Interpreting Attention Models with Human Visual Attention in Machine Reading Comprehension
This work addresses the interpretability of attention mechanisms in NLP for researchers, showing that similarity to human attention does not ensure optimal performance, which is an incremental insight.
The paper investigated the relationship between human visual attention and neural attention in machine reading comprehension by comparing LSTM, CNN, and XLNet models using a new eye-tracking dataset, finding that higher similarity to human attention correlated with LSTM and CNN models but not with the best-performing XLNet.
While neural networks with attention mechanisms have achieved superior performance on many natural language processing tasks, it remains unclear to which extent learned attention resembles human visual attention. In this paper, we propose a new method that leverages eye-tracking data to investigate the relationship between human visual attention and neural attention in machine reading comprehension. To this end, we introduce a novel 23 participant eye tracking dataset - MQA-RC, in which participants read movie plots and answered pre-defined questions. We compare state of the art networks based on long short-term memory (LSTM), convolutional neural models (CNN) and XLNet Transformer architectures. We find that higher similarity to human attention and performance significantly correlates to the LSTM and CNN models. However, we show this relationship does not hold true for the XLNet models -- despite the fact that the XLNet performs best on this challenging task. Our results suggest that different architectures seem to learn rather different neural attention strategies and similarity of neural to human attention does not guarantee best performance.