CFE 2020: Start Registration
View Submission - CMStatistics
B1165
Title: Applications of machine learning in data and text Authors:  Hongyan Cui - Beijing University of Posts and Telecommunications (China) [presenting]
Ensen Wu - Beijing University of Posts and Telecommunications (China)
Zhenwei Sun - Beijing University of Posts and Telecommunications (China)
Gangkun Wang - Beijing University of Posts and Telecommunications (China)
Songdeng Hui - Beijing University of Posts and Telecommunications (China)
Abstract: In recent decades, machine learning has played a significant role in a wide spectrum of industries due to its excellent processing ability. According to different types of data and tasks, machine learning applications can be divided into different domains, such as structure data classification, NL2SQL tasks and Semi-supervised short text classification etc. However, there are still many problems in these domains that need to be further explored, such as imbalanced classification problem, multi-task learning problems and sparsity and limited labeled data problems. Examples of our recent researches in machine learning applications in the above domains are described. Firstly, we will introduce the application of Generative Adversarial Networks and Autoencoders to solve imbalanced classification problem in structure data classification. Secondly, we will describe a pre-trained language model structure that performs well in NL2SQL tasks. Finally, we will introduce a graph neural network and self-training method, which improves the classification performance of short text.