# Language Model Pre-training for Hierarchical Document Representations

arXiv: https://arxiv.org/abs/1901.09128 · HTML: https://mlsanity.com/paper/1901.09128/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Jan 2019.

## Built on by (most cited first, 7 of 7)

- [ReadTwice: Reading Very Large Documents with Memories](https://mlsanity.com/paper/2105.04241/) (2021)
- [Iterative Hierarchical Attention for Answering Complex Questions over Long Documents](https://mlsanity.com/paper/2106.00200/) (2021)
- [Pre-training Transformer Models with Sentence-Level Objectives for Answer Sentence Selection](https://mlsanity.com/paper/2205.10455/) (2022)
- [Hierarchical Neural Network Approaches for Long Document Classification](https://mlsanity.com/paper/2201.06774/) (2022)
- [Segmenting Messy Text: Detecting Boundaries in Text Derived from Historical Newspaper Images](https://mlsanity.com/paper/2312.12773/) (2023)
- [Word Embedding Techniques for Classification of Star Ratings](https://mlsanity.com/paper/2504.13653/) (2025)
- [Enhancing Science Classroom Discourse Analysis through Joint Multi-Task Learning for Reasoning-Component Classification](https://mlsanity.com/paper/2604.21137/) (2026)

## Abstract (arXiv)

Hierarchical neural architectures are often used to capture long-distance dependencies and have been applied to many document-level tasks such as summarization, document segmentation, and sentiment analysis. However, effective usage of such a large context can be difficult to learn, especially in the case where there is limited labeled data available. Building on the recent success of language model pretraining methods for learning flat representations of text, we propose algorithms for pre-training hierarchical document representations from unlabeled data. Unlike prior work, which has focused on pre-training contextual token representations or context-independent {sentence/paragraph} representations, our hierarchical document representations include fixed-length sentence/paragraph representations which integrate contextual information from the entire documents. Experiments on document segmentation, document-level question answering, and extractive document summarization demonstrate the effectiveness of the proposed pre-training algorithms.