Over the years, cybercriminals increasingly joined the underground economy to exchange malicious services for conducting data breaches crimes. As many service providers are rippers, most cybercriminals rely on a few high quality services. To this end, cybercriminals post customer reviews evaluating the purchase experience and the service quality. To identify high quality services, researchers face two major challenges - the cybercriminal-specific language and the scale of the underground economy. This study presents a text-mining-based system for identifying high quality services by analyzing customer reviews. A novel supervised topic model is designed to accommodate the heterogeneous and uncertain nature of customer reviews. We further designed a variational algorithm for model inference. Moreover, we collected real data from two underground economy forums for English-speaking and Russian-speaking cybercriminals as our research testbed. Our research contributes to the practice of understanding and mitigating underground economy by providing cybersecurity researchers and practitioners with actionable intelligence.