基于深度学习的财务机器人自动撰文场景研究

Research on Automatic Writing Scenario of Financial Robot Based on Deep Learning

文本自动生成旨在使机器人如同人一样思考并自动撰文,减少工作人员的工作量、提高工作效率。文章将自动撰文场景建模成财务机器人自动撰写报告的问题模型,提出了基于深度学习的LSTM和GRU正反序列两种算法来实现文本自动生成。实验表明,在大规模语料库基础上该模型可取得较好的效果,该模型生成的文本所涵盖的信息更加全面,机器生成的文本与人工生成的文本意思更相近。

Automatic text generation aims to make the robot think and write automatically like human,so as to reduce the workload of the staff and improve work efficiency.In this paper,the automatic writing scene is modeled as the problem model of financial robot automatically writing reports,and an automatic text generation algorithm based on deep learning is proposed.Compared with LSTM algorithm and GRU positive and negative sequence algorithm,the experiment shows that this model can achieve good results on the basis of large-scale corpus,the text generated by this model is more comprehensive,and the meaning of the machine-generated text is more similar to that of the manually generated text.