ACL 2021long77 citations

What Context Features Can Transformer Language Models Use?

Joe O’Connor, Jacob Andreas

Abstract

Transformer-based language models benefit from conditioning on contexts of hundreds to thousands of previous tokens. What aspects of these contexts contribute to accurate model prediction? We describe a series of experiments that measure usable information by selectively ablating lexical and structural information in transformer language models trained on English Wikipedia. In both mid- and long-range contexts, we find that several extremely destructive context manipulations—including shuffling word order within sentences and deleting all words other than nouns—remove less than 15% of the usable information. Our results suggest that long contexts, but not their detailed syntactic and propositional content, are important for the low perplexity of current transformer language models.

BibTeX
@inproceedings{oconnor-andreas-2021-context,
    title = "What Context Features Can Transformer Language Models Use?",
    author = "O{'}Connor, Joe  and
      Andreas, Jacob",
    editor = "Zong, Chengqing  and
      Xia, Fei  and
      Li, Wenjie  and
      Navigli, Roberto",
    booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.acl-long.70/",
    doi = "10.18653/v1/2021.acl-long.70",
    pages = "851--864"
}
What Context Features Can Transformer Language Models Use? · ACL 2021