Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors

TL;DR

This paper introduces a dictionary learning-based visualization method to analyze transformer models, revealing hierarchical semantic structures and providing new insights into their internal representations.

Contribution

It presents a novel visualization approach that models transformer embeddings as linear superpositions of factors, enhancing interpretability of transformer representations.

Findings

Hierarchical semantic structures are captured by transformer factors

Word-level polysemy disambiguation is visualized

Long-range dependencies are identified and analyzed

Abstract

Transformer networks have revolutionized NLP representation learning since they were introduced. Though a great effort has been made to explain the representation in transformers, it is widely recognized that our understanding is not sufficient. One important reason is that there lack enough visualization tools for detailed analysis. In this paper, we propose to use dictionary learning to open up these "black boxes" as linear superpositions of transformer factors. Through visualization, we demonstrate the hierarchical semantic structures captured by the transformer factors, e.g., word-level polysemy disambiguation, sentence-level pattern formation, and long-range dependency. While some of these patterns confirm the conventional prior linguistic knowledge, the rest are relatively unexpected, which may provide new insights. We hope this visualization tool can bring further knowledge and a better understanding of how transformer networks work. The code is available at https://github.com/zeyuyun1/TransformerVis