fnctId=thesis,fnctNo=295
SNA 기법에 근거한 검색어 추천 서비스에 대한 연구
- 링크
- http://www.riss.kr/link?id=T12416767
- 작성자
- 조지은
- 저자
- 위성광
- 발행사항
- 부산 : 부산대학교, 2011
- 발행일
- 2011.02
- 국문초록
- 영문초록
- In knowledge societies, knowledge is the basis of all human activity and has also become the major creative, economic, social, cultural force. All other human activities have become more and more relent on knowledge and information. Thanks to the development of modern science and technolgyoogy, human activities are no longer limited by geography because technologyogy enables us to share, archive and retrieve information.
This experimental analysis was performed on twelve months of search log files sample from Korea. NDSL(National Digital Science library)
Scholar-the major Korean science portal, from 15th April 2008 to 31st 16 5 May 2009.
Power law is a very important theory in solving and explaining chaos in the real-world. in this paper, we made two communities -user to user and query term to query term. By analyzing the properties of user and query term network, we obtained that both of the networks obey the power laws, in other words, we claimed that the internal network connection of user and query term is very complex and without order, its attribution indicates that it is a complex network.
Based on the value of phase transition point, we chose the network that contains this point of phase transition to replace the full networks.Zipf, as one kind of small world phenomenon, allows NDSL search engine to find out the main 732 users. Then, after using the cluster
analysis, the nodes of the user threshold networks can be gathered under 29 sub-groups or sub-networks based on the link information we found out that keywords or subjects related to user's information requirements. Finally, we chose the cluster-22 as the analysis target in our test.
This paper used the social network analysis to analyze a various attributes of searchers "searching behaviors that appeared in search access log data. Based on this theory, we recommended the related list about the user query by its linkage through analyzing the Ego-network. Then we compared to the similarity coefficient results between Social Network Analysis method and Apriori algorithm of Mining.
Results are as follows. First, the structure of network relays on the similarity of the queries which users inputted. Some queries were shared with ego-searcher and alter searchers. Second, the total number of searchers can be divided into sub-groups by using the clustering analysis. Each sub-group has its own subjects. Third, compared to the results of two tests, Social Network Analysis method is more advanced qualitatively than Apriori algorithm of mining method. This study reveals a new recommendation algorithm bases on the search query through the social network analysis, it is capable of being applied.
However, we only focused on the search query to analyze the network Afterwards, in order to provide more accurate, effective personal information service in digital library, our study should take other factors into account, such as click log file, download log file etc. Moreover, in the process of morpheme analysis and Cluster analysis, there are a lot of questions to be deeply studied.
- 일반텍스트
- 첨부파일