Companies and organizations equipped with IT infrastructure usually face security threats due to vulnerabilities in information systems. This paper aims to build models using intelligent algorithms to automatically identify vulnerability types and predict risk levels. We first collect reports from a Chinese vulnerability crowd-testing platform, then establish models by using textual representation technologies, shallow and deep learning algorithms. The experimental results show that the deep learning model with neural text representation could achieve better performance of vulnerability identification and risk level prediction. This research contributes to the information security literature and could help companies and organizations to more efficiently fix information systems vulnerabilities.