In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6907–6919, Dublin, Ireland.

Word Order Does Matter and Shuffled Language Models Know It.

Anthology ID: 2022.acl-long.476 Volume: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) Month: May Year: 2022 Address: Dublin, Ireland Venue: ACL SIG: Publisher: Association for Computational Linguistics Note: Pages: 6907–6919 Language: URL: DOI: 10.18653/v1/2022.acl-long.476 Bibkey: abdou-etal-2022-word Cite (ACL): Mostafa Abdou, Vinit Ravishankar, Artur Kulmizev, and Anders Søgaard. Finally, we show that beyond GLUE, a variety of language understanding tasks do require word order information, often to an extent that cannot be learned through fine-tuning. Surprisingly, we find even Language models trained on text shuffled after subword segmentation retain some semblance of information about word order because of the statistical dependencies between sentence length and unigram probabilities. We show this is in part due to a subtlety in how shuffling is implemented in previous work – before rather than after subword segmentation. We probe these language models for word order information and investigate what position embeddings learned from shuffled text encode, showing that these models retain a notion of word order information. Somewhat counter-intuitively, some of these studies also report that position embeddings appear to be crucial for models’ good performance with shuffled text. Abstract Recent studies have shown that language models pretrained and/or fine-tuned on randomly permuted sentences exhibit competitive performance on GLUE, putting into question the importance of word order information.

Semblance of order download