A Novel Science Literature Search Engine

Modern researchers share and find research material conveniently for the development of science and Internet. It is a terrible experience scholar will face a long and chaos paper list when they use search engines to search science literature they are not familiar with in a huge database. It is not easy to gain insights from this organization of traditional science literature search engine. This thesis analysis the strategy of traditional science literature search engine and try to improve the experience when user is not familiar with the field they search. Some new algorithm was designed to achieve this goal. These new algorithm include a new simple and efficient key phrase extraction algorithm, a PageRank based science literature clustering algorithm, a novel fast MPM training algorithm. A small experimental science literature search engine based on these new algorithms was build. The experiment shows that those new algorithm did improve the search experience when a user search a field that he/she is not familiar with.