The Self-adaptive Crowdsourcing Based On Users’ Behaviors Analyses In Digital Library

With the rapid development of Internet, an increasingly amount of information need to be stored and spread in people’s daily lives, the form and feature of the information become more and more diversified. The mode of traditional library cannot afford to satisfy with people’s requirements in the era of information, so the digital library emerges as required. The digital library is the development and continuation of traditional library, which can store lots of digital resources, with a low cost to maintain, hard to be damaged and easy to be searched. The users can have easy access to these digital resources via the Internet conveniently.The technology of crowdsourcing, which is one kind of the new production organizations, is also brought about by the technology of the Internet. It splits the tasks that not easily done by computers into small ones and distribute them to different workers to finish, and it promises to pay the workers back for their contributions. In such way it can make use of the free time of people, improve the efficiency of the workers, and lower the cost at the same time.This thesis involves with digital library and the technology of crowdsourcing, the main work of this thesis is as follows:First, the paper come up with the method of self-adaptive crowdsourcing based on the analyses of users’ behaviors under the circumstances of digital library. At first, by analyzing the users’behaviors, we forecast the crowdsourcing workers from users, to decrease the blindness of selecting users randomly. Then, we take advantage of the model of Multi-armed Bandit to select crowdsourcing workers self adaptively, for the purpose of finishing the crowdsourcing work efficiently.Second, it is accomplished that the self-adaptive crowdsourcing system based on the analyses of users’ behaviors in the platform of CADAL. The system has several functional modules that include potential crowdsourcing workers’ forecast, mails’ invitation, feedbacks’ collection and crowdsourcing workers’ selection based on Multi-armed Bandit. We have experiment several methods of selecting crowdsourcing workers, and the experiment results are analysed.In a word, the thesis is to study the methods of how to apply the technology of crowdsourcing to digital library. What’s more, we propose the method that by analyzing the users’ behaviors, and combines the model of Multi-armed bandit with recommender system to improve the performance of crowdsourcing in an innovative way. We also implement the system of self-adaptive crowdsourcing on the platform of CADAL. The method has a good effect on helping to solve the problem of the deficiency of the metadata information in the journal of republic of China.