NextLevelBERT: Masked Language Modeling with Higher-Level Representations for Long Documents

Tamara Czinczoll, Christoph Hönes, Maximilian Schall, Gerard De Melo


Abstract
While (large) language models have significantly improved over the last years, they still struggle to sensibly process long sequences found, e.g., in books, due to the quadratic scaling of the underlying attention mechanism. To address this, we propose NextLevelBERT, a Masked Language Model operating not on tokens, but on higher-level semantic representations in the form of text embeddings. We pretrain NextLevelBERT to predict the vector representation of entire masked text chunks and evaluate the effectiveness of the resulting document vectors on three types of tasks: 1) Semantic Textual Similarity via zero-shot document embeddings, 2) Long document classification, 3) Multiple-choice question answering. We find that next-level Masked Language Modeling is an effective technique to tackle long-document use cases and can outperform much larger embedding models as long as the required level of detail of semantic information is not too fine. Our models and code are publicly available online.
Anthology ID:
2024.acl-long.256
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
4656–4666
Language:
URL:
https://aclanthology.org/2024.acl-long.256
DOI:
10.18653/v1/2024.acl-long.256
Bibkey:
Cite (ACL):
Tamara Czinczoll, Christoph Hönes, Maximilian Schall, and Gerard De Melo. 2024. NextLevelBERT: Masked Language Modeling with Higher-Level Representations for Long Documents. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 4656–4666, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
NextLevelBERT: Masked Language Modeling with Higher-Level Representations for Long Documents (Czinczoll et al., ACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.acl-long.256.pdf