Large general-purposed community question-answering sites are becoming popular as a new venue for generating knowledge and helping users in their information needs. In this paper we analyze the characteristics of knowledge generation and user participation behavior in the largest question-answering online community in South Korea, Naver Knowledge-iN. We collected and analyzed over 2.6 million question/answer pairs from fifteen categories between 2002 and 2007, and have interviewed twenty six users to gain insights into their motivations,roles, usage and expertise. We find altruism, learning, and competency are frequent motivations for top answerers to participate, but that participation is often highly intermittent. Using a simple measure of user performance, we find that higher levels of participation correlate with better performance. We also observe that users are motivated in part through a point system to build a comprehensive knowledge database. These and other insights have significant implications for future knowledge generating online communities.

Online discussions such as a large-scale community brainstorming often end up with an unorganized bramble of ideas and topics that are difficult to reuse. A process of distillation is needed to boil down a large information space into information that is concise and organized. We take a system-augmented approach to the problem by creating a set of tools with which human editors can collaboratively distill a large amount of informal information.

Help-seeking communities have been playing an increasingly critical role in the way people seek and share information. However, traditional help-seeking mechanisms of these online communities have some limitations. In this paper, we describe an expertise-finding mechanism that attempts to alleviate the limitations caused by not knowing users' expertise levels. As a result of using social network data from the online community, this mechanism can automatically infer expertise level. This allows, for example, a question list to be personalized to the user's expertise level as well as to keyword similarity. We believe this expertise location mechanism will facilitate the development of next generation help-seeking communities.