Comparing Attention-Based Convolutional and Recurrent Neural Networks: Success and Limitations in Machine Reading Comprehension
Matthias Blohm, Glorianna Jagfeld, Ekta Sood, Xiang Yu, Ngoc Thang Vu
Proc. ACL SIGNLL Conference on Computational Natural Language Learning (CoNLL), pp. 108–118, 2018.
Abstract
We propose a machine reading comprehension model based on the compare-aggregate framework with two-staged attention that achieves state-of-the-art results on the MovieQA question answering dataset. To investigate the limitations of our model as well as the behavioral difference between convolutional and recurrent neural networks, we generate adversarial examples to confuse the model and compare to human performance. Furthermore, we assess the generalizability of our model by analyzing its differences to human inference, drawing upon insights from cognitive science.Links
doi: 10.18653/v1/K18-1011
Paper: blohm18_conll.pdf
Paper Access: https://www.aclweb.org/anthology/K18-1011
BibTeX
@inproceedings{blohm18_conll,
title = {Comparing Attention-Based Convolutional and Recurrent Neural Networks: Success and Limitations in Machine Reading Comprehension},
author = {Blohm, Matthias and Jagfeld, Glorianna and Sood, Ekta and Yu, Xiang and Vu, Ngoc Thang},
booktitle = {Proc. ACL SIGNLL Conference on Computational Natural Language Learning (CoNLL)},
year = {2018},
publisher = {Association for Computational Linguistics},
url = {https://www.aclweb.org/anthology/K18-1011},
doi = {10.18653/v1/K18-1011},
pages = {108--118}
}