| Since the birth of library over 3000 years ago,its functions have been centering on manual service for bamboo or paper books,which were rather inefficient.The progress of Information Technology in the recent 30 years(characterized by PC and Internet),however,has provided the technology required for digital library.Since DL1 in 1990s in the US.digital library has been evolving extremely fast.Digitalization of paper books at the beginning,all the way to Google's ambition of Virtual Library,these developments in less than two decades have made great impact on library,a facility of thousands of years old.All countries and regions have invested heavily on R&D and establishment of digital libraries,regarding this as a country's information infrastructure.Starting in 1996.demonstrative projects such as CPDLP.CDLP,NSTL.CALIS have facilitated the progress of digital library in China.In 1996,S.Sutton.a British scholar,put forward the concept of Hybrid Library-not a traditional paper-based library,nor a purely digital library,but a combination of both,and suggested that such hybrid libraries will dominate in long time to come.Hybrid libraries will play the parts of both traditional libraries and digital libraries.While libraries are being digitalized.huge amounts of data are generated,and the technology of data mining,which came into being at about the same time as digital library,can serve as a best tool to analyze such huge amounts of data.Many researchers home and abroad have advanced analyses and applications on relevant data in various ways,but few have provided a comprehensive study of university hybrid libraries.In this paper,we use data mining to analyze real-time data,revealing the working of hybrid libraries and optimize their processes.We discuss key parts in the working of hybrid libraries,such as resource purchase model,catalog atomization and personalized services based on users' interests.Namely.1)We analyze the library of Nanjing University of Finance and Economics and advance a model for hybrid library,combining our daily application and the establishing of a hybrid library.We point out that process optimization can utilize resource acquisition,improve working environment,reduce management costs and upgrade the library's services.2)For book purchase.we mine the data of readers' borrowing behavior,which exhibit the uses of paper books,to build a purchase model,thus reducing copies and purchases while still meeting the readers' demand.The saved fund can go to the establishment of digital resources,Moreover,we construct the secondar categories in the library and test their feasibility,cutting down the categorization process and thus reducing costs.3)We introduce Information Automation Technology into cataloging,i.e.,books automatically cataloged via machine learning,which overcomes problems caused by knowledge engineering or data mining in cataloging.We then test the precision and rationale of cataloging by machine learning.We also test the automatic indexing based on Conditional Random Fields(CRF),showing that under such conditions indexing can happen through incomplete book titles.Meanwhile,we analyze readers' borrowing behavior through mining 10 years of data,including readers' reading interests,interest groups,and correlation between course grades and reading habits.All this will contribute to the establishment of personalized service of the library,In this paper,our contributions are 1)a purchase model via cluster analysis and statistical analysis is constructed and can apply to real work,while the identification of core users,relationship between readers' majors and borrowing habits,and correlation between course grades and borrowing habits will improve the relevance of library services to each individual user;2)the process optimization can improve resource acquisition,reducing costs and potential errors through cataloging and other processes;and 3)we show that machine learning can be applied to cataloging and that automatic indexing based on CRF is more simplified. |
PDF Full Text Request