Crowdsourcing transcription projects in digital humanity domain are of great significance to the transcription, storage, disseminate and development of cultural heritage. However, because of the heavy workload and the high-level professionalism in the crowdsourcing projects, it becomes necessary to ensure the performance of crowdsourcing tasks. This paper adopts Transcribe-Sheng Project, one of the typical crowdsourcing projects in digital humanity domain in China, as a case. We conduct a quasi-experiment to explore the difference of task performance in the crowdsourcing transcription projects from two perspectives-task complexity and domain knowledge. Our findings reveal that both task complexity and domain knowledge significantly influence task performance. What's more, there is an interaction effect exists between task complexity and domain knowledge on task performance. Based on these findings, we provide some practical suggestions on task design, participants' recruitment and task fit for the crowdsourcing transcription projects in digital humanity domain in China.