Knowledge Sharing via Social Login: Exploiting Microblogging Service for Warming up Social Question Answering Websites

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 656–666, Dublin, Ireland, August 23-29 2014. Knowledge Sharing via Social Login: Exploiting Microblogging Service for Warming up Social Question Answering Websites Yang Xiao1, Wayne Xin Zhao2, Kun Wang1 and Zhen Xiao1 1School of Electronics Engineering and Computer Science, Peking University, China 2School of Information, Renmin University of China, China {xiaoyangpku, batmanfly}@gmail.com {wangkun, xiaozhen}@net.pku.edu.cn Abstract Community Question Answering (CQA) websites such as Quora are widely used for users to get high quality answers. Users are the most important resource for CQA services, and the awareness of user expertise at early stage is critical to improve user experience and reduce churn rate. However, due to the lack of engagement, it is difficult to infer the expertise levels of newcomers. Despite that newcomers expose little expertise evidence in CQA services, they might have left footprints on external social media websites. Social login is a technical mechanism to unify multiple social identities on different sites corresponding to a single person entity. We utilize the social login as a bridge and leverage social media knowledge for improving user performance prediction in CQA services. In this paper, we construct a dataset of 20,742 users who have been linked across Zhihu (similar to Quora) and Sina Weibo. We perform extensive experiments including hypothesis test and real task evaluation. The results of hypothesis test indicate that both prestige and relevance knowledge on Weibo are correlated with user performance in Zhihu. The evaluation results suggest that the social media knowledge largely improves the performance when the available training data is not sufficient. 1 Introduction One of the main challenges for social startup websites is how to gain a considerable number of users quickly. A growing number of social startups outsource sign-up process to existing social networking services. They allow users to log in to the services using their existing social media accounts. For exam- ple, Quora allows users to log in with their Google, Twitter or Facebook accounts based on the OpenID technology. Lots of startup web services benefit from the huge number of users and rich relationships accumulated by social network sites. Social login helps the newborn web services to collect crowds of users in a short time. Moreover, startup web services can gain reliable profiles through social login. It also offers a convenient mechanism for users to surf the web using a unified social identity (e.g., Twit- ter account). For example, by the end of 2013, there are about 600,000 web services including mobile applications using social login offered by Sina Weibo. When we go beyond simple import of profiles and consider the general problem of leveraging knowl- edge from social media, many subtasks arise. One of them is how to incorporate data from social media and startup web service to better predict user performance. In this paper, we take the largest social based question answering service Zhihu in China, which closely resembles Quora, as the testbed. Different from traditional CQA sites such …

Content truncated to 500 words