arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2001.04063cs.CL
收藏

ProphetNet:预测未来N-gram用于序列到序列预训练

ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training

University of Science and Technology of China(中国科学技术大学);Microsoft(微软);Microsoft Research Asia(微软亚洲研究院);Sichuan University(四川大学)

Weizhen Qi, Yu Yan, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang, Ming Zhou

摘要:

本文提出了一种新的序列到序列的预训练模型,称为神经网络,它引入了一种新的自我监督的目标命名为未来的n-gram预测和建议的n-stream自我注意力机制。与传统的序列到序列模型中的提前一步预测不同,该算法通过提前n步预测来优化,该预测基于每个时间步的先前上下文令牌同时预测接下来的n个令牌。未来的n-gram预测明确地鼓励模型为未来的token做计划,并防止在强局部相关性上的过拟合。我们分别使用基本规模数据集(16 GB)和大规模数据集(160 GB)预训练AptNet。然后,我们在CNN/DailyMail,Gigaword和SQuAD 1.1基准测试中进行了抽象摘要和问题生成任务的实验。实验结果表明,与使用相同规模的预训练语料库的模型相比,AptNet在所有这些数据集上都取得了新的最先进的结果。

英文摘要:

This paper presents a new sequence-to-sequence pre-training model called ProphetNet, which introduces a novel self-supervised objective named future n-gram prediction and the proposed n-stream self-attention mechanism. Instead of optimizing one-step-ahead prediction in the traditional sequence-to-sequence model, the ProphetNet is optimized by n-step ahead prediction that predicts the next n tokens simultaneously based on previous context tokens at each time step. The future n-gram prediction explicitly encourages the model to plan for the future tokens and prevent overfitting on strong local correlations. We pre-train ProphetNet using a base scale dataset (16GB) and a large-scale dataset (160GB), respectively. Then we conduct experiments on CNN/DailyMail, Gigaword, and SQuAD 1.1 benchmarks for abstractive summarization and question generation tasks. Experimental results show that ProphetNet achieves new state-of-the-art results on all these datasets compared to the models using the same scale pre-training corpus.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑