基于BERT模型的智能数据分析技术

Intelligent Data Analysis Technology Based on BERT Model

面对日趋增长的数据分析需求,以人工编写SQL方式进行数据分析已无法满足要求,而基于自然语言交互界面的数据分析已成为发展趋势。文章提出了一种基于来自变换器的双向编码器表征量(Bidirectional Encoder Representations from Transformers,BERT)模型的智能数据分析技术,相对于Word2Vec/全局唯一标识分区表(Globally Unique Identifier Partition Table,GPT)等模型,大幅提升了自然语言到SQL转换的准确率,使自然语言交互式数据分析准确率超过人工编写SQL的方式。

In the face of the growing demand for data analysis, the efficiency of manually writing SQL for data analysis can no longer meet the requirements, and data analysis based on natural language interaction interface has become a development trend. This paper proposes an intelligent data analysis technology based on Bidirectional Encoder Representations from Transformers(BERT) model. Compared with Word2Vec/Globally Unique Identifier Partition Table(GPT) and other models, it greatly improves the accuracy of natural language to SQL conversion, and makes the accuracy of natural language interactive data analysis exceed that of manually written SQL.