arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.15871cs.CLcs.LG
收藏

利用海量金融语料提高NER性能

Improving NER's Performance with Massive financial corpus

Han Zhang

摘要:

训练大型深度神经网络需要大量高质量的标注数据,但时间和人力成本对于小型企业来说过于昂贵。我们从一个小规模、低质量的训练数据开始进行公司名称识别任务,然后使用技巧来提高模型的训练速度,并以最小的人力成本来预测性能。我们使用的方法包括在金融语料库中预训练一个简单的语言模型,如Albert-small或Electra-small,知识的提取和多阶段学习。实验结果表明,该模型的召回率提高了近20个百分点,比BERT-CRF模型快4倍。

英文摘要:

Training large deep neural networks needs massive high quality annotation data, but the time and labor costs are too expensive for small business. We start a company-name recognition task with a small scale and low quality training data, then using skills to enhanced model training speed and predicting performance with minimum labor cost. The methods we use involve pre-training a lite language model such as Albert-small or Electra-small in financial corpus, knowledge of distillation and multi-stage learning. The result is that we raised the recall rate by nearly 20 points and get 4 times as fast as BERT-CRF model.

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。

↑