As an efficient online academic information repository and information channel with crowds’ contribution, online research social platforms have become an efficient tool for various kinds of research & management applications. Social network platforms have also become a major source to seek for field experts. They have advantages of crowd contributions, easy to access without geographic restrictions and avoiding conflict of interests over traditional database and search engine based approaches. However, current research attempts to find experts based on features such as published research work, social relationships, and online behaviours (e.g. reads and downloads of publications) on social platforms, they ignore to verify the reliability of identified experts. To bridge this gap, this research proposes an innovative Topic Sensitive SimRank (TSSR) model to identify “real” experts on social network platforms. TSSR model includes three components: LDA for Expertise Extension, Topic Sensitive Network for Reputation Measurement, and Topic Sensitive SimRank for unsuitable experts detection. We also design a parallel computing strategy to improve the efficiency of the proposed methods. Last, to verify the effectiveness of the proposed model, we design an experiment on one of the research social platforms-ScholarMate to seek for experts for companies that need academic-industry collaboration.