Abstract: ［Purpose/Significance］The advent of live knowledge broadcasting platforms has brought users a new approach of knowledge acquisition, communication and monetization. Research on product feature extraction and problem mining for knowledge broadcasting is helpful to improve the quality of knowledge broadcasting products and promote a benign development of knowledge broadcasting platforms.［Design/Methodology］Taking Zhihu Live as the research object, this paper extracted the high-frequency words of product features from 9,108 users’comments so as to build a feature system of knowledge broadcasting products; with the use of fine-grained viewpoint extraction methods, viewpoints of negative sentiment were extracted in order to reveal the existing problems of knowledge live broadcasting products.［Findings/Conclusion］The product feature system of Zhihu Live lectures can be summarized into three dimensions, namely lecturers, courses and users. Among the three dimensions, users are most concerned with features related to the dimension of courses such as the quality and the form of contents. Currently, the main problems of knowledge live broadcasting products are inadequate real valuable contents in courses, unserious answering of questions, unreasonable time arrangement and so on. Consequently, platforms should collaborate with lecturers to improve service quality through problem-oriented methods.［Originality/Value］Focusing on new knowledge broadcasting approaches, this study provides a reference for improving knowledge service and users’ satisfaction of knowledge broadcasting platforms.

Keywords: Knowledge live broadcasting platform, User comment, Feature extraction, Opinion extraction, Zhihu Live