Heads-up! Unsupervised Constituency Parsing via Self-Attention Heads
Preprint
Abstract
Transformer-based pre-trained language models (PLMs) have dramatically improved the state of the art in NLP across many tasks. This has led to substantial interest in analyzing the syntactic knowledge PLMs learn. Previous approaches to this question have been limited, mostly using test suites or probes. Here, we propose a novel fully unsupervised parsing approach that extracts constituency trees from PLM attention heads. We rank transformer attention heads based on their inherent properties, and create an ensemble of high-ranking heads to produce the final tree. Our method is adaptable to low-resource languages, as it does not rely on development sets, which can be expensive to annotate. Our experiments show that the proposed method often outperform existing approaches if there is no development set present. Our unsupervised parser can also be used as a tool to analyze the grammars PLMs learn implicitly. For this, we use the parse trees induced by our method to train a neural PCFG and compare it to a grammar derived from a human-annotated treebank.Keywords
All Related Versions
This publication has 15 references indexed in Scilit:
- Information-Theoretic Probing with Minimum Description Length2020
- Analysis Methods in Neural Language Processing: A SurveyTransactions of the Association for Computational Linguistics, 2019
- A Structural Probe for Finding Syntax in Word Representations2019
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding2019
- Towards Realistic Practices In Low-Resource Natural Language Processing: The Development Set2019
- A Multiscale Visualization of Attention in the Transformer Model2019
- Linguistic Knowledge and Transferability of Contextual Representations2019
- Unsupervised Recurrent Neural Network Grammars2019
- Unsupervised Latent Tree Induction with Deep Inside-Outside Recursive Auto-Encoders2019
- Unsupervised Learning of Syntactic Structure with Invertible Neural Projections2018