arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2011.02610cs.CLcs.AI
收藏

通过时间感知预训练改进事件持续时间预测

Improving Event Duration Prediction via Time-aware Pre-training

Cornell University(康奈尔大学)

Zonglin Yang, Xinya Du, Alexander Rush, Claire Cardie

摘要:

NLP中的端到端模型很少编码关于时间长度的外部世界知识。我们介绍了两种有效的持续时间预测模型,通过阅读时间相关的新闻句子(时间感知预训练)来结合外部知识。具体来说,一个模型预测的范围/单位,其中的持续时间值落在(R-pred);和其他预测准确的持续时间值E-pred。我们最好的模型- E-pred,大大优于以前的工作,并捕捉持续时间信息更准确地比R-pred。我们还证明了我们的模型能够在无监督设置持续时间预测,优于基线。

英文摘要:

End-to-end models in NLP rarely encode external world knowledge about length of time. We introduce two effective models for duration prediction, which incorporate external knowledge by reading temporal-related news sentences (time-aware pre-training). Specifically, one model predicts the range/unit where the duration value falls in (R-pred); and the other predicts the exact duration value E-pred. Our best model -- E-pred, substantially outperforms previous work, and captures duration information more accurately than R-pred. We also demonstrate our models are capable of duration prediction in the unsupervised setting, outperforming the baselines.

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。

↑