User Incentive Mechanism For Collaborative Tagging Quality

Collaborative tagging system are the platforms which allow users in different places to annotate resources online. This is a representative application of network resource management that takes advantage of crowdsourcing mechanism. Tagging data plays an important role in area of massive web resource searching, mining, and recommending. By using the tagging data, social tagging system can simply manage resources. However, tagging data always have low quality problems in real application, such as containing irrelevant information, misspelling, synonyms, polysemy, and so on. These noisy and incomplete tags severely affect the description of marked resources, which has become a bottleneck of tag-based applications. Therefore, how to improve the quality of tags has become a hot-spot issue in area of collaborative tagging.Three kinds of methods are discussed in solving the problem of low quality tags. They are tag recommendation, semantic based tagging and incentive-based sufficient tagging, respectively. However, tag recommendation restricts the users’creativity and thus fades the natural of crowdsourcing. And, the implementation procedure of semantic based tagging is always complicated and thus increases the burden of users in some extent. Incentive-based sufficient tagging is a natural tagging method. Studies have shown that, if a resource receives enough number of posts, its tagging state will reach stable point, and these tagging data can make the resource more accuracy. Although the existing incentive-based tagging mechanism can stimulate users to annotate under-tagged resources so as to reach the stable point, it can’t distinguish posts with different quality. In response to these problem, we propose a post quality based dynamic incentive mechanism (PQIM) and the implementation plan. The details are as follows:Quantitative methods for a user post are proposed. According to the statistical results of resource’s historical tagging data, we analyze the characteristics of resource’s relative stable tag set, and give the standards for metric of stable tag set. Based on the resource’s relative stable tag set, we design four kinds of metric methods from aspect of tag multiplicity, frequency, size of a post and resource stability, respectively.A post quality based dynamic incentive mechanism is proposed. We set incentive rules from users’tagging time and post quality. The earlier a user tagging a resource, the higher award the user receive; the higher quality a user post have, the more award the user get. And, we design the dynamic incentive function based on post quantitative method. Making award negative correlation with resource’s tagging state and positive correlation with user’s post-quality. According to the mechanism design theory in game theory, we theoretically verify the effectiveness of our post quality based dynamic incentive mechanism.The system architecture of PQIM and its implementation algorithm are proposed. Based on the real datasets, we analyze the effectiveness of our mechanism from the improvement of system tagging quality and user payoff respectively. For the improvement of system tagging quality, we propose the highest reward priority strategy corresponding to our mechanism. By comparing to the best strategy corresponding to the existing mechanism, our method can not only have better system quality under given budget, but also save time in the process of achieving desired system quality. For user payoffs, we firstly make statisticanalysis on history data, and extract active users to analyze their payoff when tagging in different time slot and with different post quality under PQIM. The results verify our mechanism’s incentive on users.