# Addressing the Length Bias Problem in Document-Level Neural Machine Translation

arXiv: https://arxiv.org/abs/2311.11601 · HTML: https://mlsanity.com/paper/2311.11601/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 43 references that are in the map.
- First posted on arXiv 20 Nov 2023.

## Builds on (most cited first, 22 of 22)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation](https://mlsanity.com/paper/2108.12409/) (2021)
- [Multilingual Denoising Pre-training for Neural Machine Translation](https://mlsanity.com/paper/2001.08210/) (2020)
- [From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification](https://mlsanity.com/paper/1602.02068/) (2016)
- [Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges](https://mlsanity.com/paper/1907.05019/) (2019)
- [Randomized Positional Encodings Boost Length Generalization of Transformers](https://mlsanity.com/paper/2305.16843/) (2023)
- [Overcoming a Theoretical Limitation of Self-Attention](https://mlsanity.com/paper/2202.12172/) (2022)
- [Neural Machine Translation with Extended Context](https://mlsanity.com/paper/1708.05943/) (2017)
- [KERPLE: Kernelized Relative Positional Embedding for Length Extrapolation](https://mlsanity.com/paper/2205.09921/) (2022)
- [Document-Level Neural Machine Translation with Hierarchical Attention Networks](https://mlsanity.com/paper/1809.01576/) (2018)
- [Improving the Transformer Translation Model with Document-Level Context](https://mlsanity.com/paper/1810.03581/) (2018)
- [Selective Attention for Context-aware Neural Machine Translation](https://mlsanity.com/paper/1903.08788/) (2019)
- [Microsoft Translator at WMT 2019: Towards Large-Scale Document-Level Neural Machine Translation](https://mlsanity.com/paper/1907.06170/) (2019)
- [Exploiting Cross-Sentence Context for Neural Machine Translation](https://mlsanity.com/paper/1704.04347/) (2017)
- [G-Transformer for Document-level Machine Translation](https://mlsanity.com/paper/2105.14761/) (2021)
- [Does Neural Machine Translation Benefit from Larger Context?](https://mlsanity.com/paper/1704.05135/) (2017)
- [Enhancing Context Modeling with a Query-Guided Capsule Network for Document-level Translation](https://mlsanity.com/paper/1909.00564/) (2019)
- [Fusing Recency into Neural Machine Translation with an Inter-Sentence Gate Model](https://mlsanity.com/paper/1806.04466/) (2018)

## Built on by (most cited first, 1 of 1)

- [Post-hoc Reward Calibration: A Case Study on Length Bias](https://mlsanity.com/paper/2409.17407/) (2024)

## Abstract (arXiv)

Document-level neural machine translation (DNMT) has shown promising results by incorporating more context information. However, this approach also introduces a length bias problem, whereby DNMT suffers from significant translation quality degradation when decoding documents that are much shorter or longer than the maximum sequence length during training. %i.e., the length bias problem. To solve the length bias problem, we propose to improve the DNMT model in training method, attention mechanism, and decoding strategy. Firstly, we propose to sample the training data dynamically to ensure a more uniform distribution across different sequence lengths. Then, we introduce a length-normalized attention mechanism to aid the model in focusing on target information, mitigating the issue of attention divergence when processing longer sequences. Lastly, we propose a sliding window strategy during decoding that integrates as much context information as possible without exceeding the maximum sequence length. The experimental results indicate that our method can bring significant improvements on several open datasets, and further analysis shows that our method can significantly alleviate the length bias problem.