Understanding Learning Dynamics Of Language Models with SVCCA

Abstract

Research has shown that neural models implicitly encode linguistic features, but there has been no research showing \emph{how} these encodings arise as the models are trained. We present the first study on the learning dynamics of neural language models, using a simple and flexible analysis method called Singular Vector Canonical Correlation Analysis (SVCCA), which enables us to compare learned representations across time and across models, without the need to evaluate directly on annotated data. We probe the evolution of syntactic, semantic, and topic representations and find that part-of-speech is learned earlier than topic; that recurrent layers become more similar to those of a tagger during training; and embedding layers less similar. Our results and methods could inform better learning algorithms for NLP models, possibly to incorporate linguistic information more effectively.

Accepted for publication in NAACL 2019

Cited by in corpus (6)

• Correlating neural and symbolic representations of language
• The Life Cycle of Knowledge in Big Language Models: A Survey
• Distilling Relation Embeddings from Pre-trained Language Models
• Multi-Task Learning with Shared Encoder for Non-Autoregressive Machine Translation
• Insights on Neural Representations for End-to-End Speech Recognition
• Is neural language acquisition similar to natural? A chronological probing study