BERT4SO: Neural Sentence Ordering by Fine-tuning BERT
arXiv:2103.13584
Abstract
Sentence ordering aims to arrange the sentences of a given text in the correct order. Recent work frames it as a ranking problem and applies deep neural networks to it. In this work, we propose a new method, named BERT4SO, by fine-tuning BERT for sentence ordering. We concatenate all sentences and compute their representations by using multiple special tokens and carefully designed segment (interval) embeddings. The tokens across multiple sentences can attend to each other which greatly enhances their interactions. We also propose a margin-based listwise ranking loss based on ListMLE to facilitate the optimization process. Experimental results on five benchmark datasets demonstrate the effectiveness of our proposed method.
References in corpus (12)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Decoupled Weight Decay Regularization
- SummaRuNNer: A Recurrent Neural Network based Sequence Model for Extractive Summarization of Documents
- Inferring Strategies for Sentence Ordering in Multidocument News Summarization
- Catching the Drift: Probabilistic Content Models, with Applications to Generation and Summarization
- A Corpus and Evaluation Framework for Deeper Understanding of Commonsense Stories
- Sentence Ordering and Coherence Modeling using Recurrent Neural Networks
- End-to-End Neural Sentence Ordering Using Pointer Network
- Neural Sentence Ordering
- Deep Ranking Model by Large Adaptive Margin Learning for Person Re-identification
- Neural Sentence Ordering Based on Constraint Graphs
- Evaluating Text Coherence at Sentence and Paragraph Levels