arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.16082cs.CL
收藏

词级神经网络探测中的量化效果度量

Measuring Memorization Effect in Word-Level Neural Networks Probing

Charles University（查理大学）

Rudolf Rosa, Tomáš Musil, David Mareček

备注：
Accepted to TSD 2020. Will be published in Springer LNCS
发表信息：
LNCS 12284, TSD (2020) 180-188
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

摘要：

多项研究探讨了在为端到端NLP任务训练的神经网络中出现的表征，并研究了哪些单词级别的语言信息可能被编码在表征中。在经典的探测中，分类器在表示上进行训练以提取目标语言信息。然而，存在分类器简单地记忆单个单词的语言标签而不是从表示中提取语言抽象的威胁，从而报告假阳性结果。虽然已经做出了相当大的努力，以尽量减少记忆的问题，实际测量的分类器中发生的记忆量的任务一直研究不足。在我们的工作中，我们提出了一种测量记忆效果的简单通用方法，该方法基于对称选择训练中可见和未可见的可比较测试单词集。我们的方法可以用来明确量化的记忆量发生在探测设置，以便可以选择一个适当的设置和探测的结果可以解释与可靠性估计。我们通过展示我们的方法在一个经过训练的神经机器翻译编码器中探测语音部分的案例研究来证明这一点。

英文摘要：

Multiple studies have probed representations emerging in neural networks trained for end-to-end NLP tasks and examined what word-level linguistic information may be encoded in the representations. In classical probing, a classifier is trained on the representations to extract the target linguistic information. However, there is a threat of the classifier simply memorizing the linguistic labels for individual words, instead of extracting the linguistic abstractions from the representations, thus reporting false positive results. While considerable efforts have been made to minimize the memorization problem, the task of actually measuring the amount of memorization happening in the classifier has been understudied so far. In our work, we propose a simple general method for measuring the memorization effect, based on a symmetric selection of comparable sets of test words seen versus unseen in training. Our method can be used to explicitly quantify the amount of memorization happening in a probing setup, so that an adequate setup can be chosen and the results of the probing can be interpreted with a reliability estimate. We exemplify this by showcasing our method on a case study of probing for part of speech in a trained neural machine translation encoder.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑