Neural language models (LMs) perform well on tasks that require sensitivity to syntactic structure. Drawing on the syntactic priming paradigm from psycholinguistics, we propose a novel technique to analyze the representations that enable such success. By establishing a gradient similarity metric between structures, this technique allows us to reconstruct the organization of the LMs' syntactic representational space. We use this technique to demonstrate that LSTM LMs' representations of different types of sentences with relative clauses are organized hierarchically in a linguistically interpretable manner, suggesting that the LMs track abstract properties of the sentence.
@article{arxiv.1909.10579,
title = {Using Priming to Uncover the Organization of Syntactic Representations in Neural Language Models},
author = {Grusha Prasad and Marten van Schijndel and Tal Linzen},
journal= {arXiv preprint arXiv:1909.10579},
year = {2019}
}
Comments
9 pages paper, 2 pages references and 3 pages supplementary materials. Code for the templates and analyses can be found here: https://github.com/grushaprasad/RNN-Priming