Automated Keyword Extraction of Learning Materials Using Semantic Relations

The poster will present our on-going research, which will develop new algorithms to automatically generate keywords from online documents that describes lesson plans in mathe- matics and science. The motivations for improving the cur- rent keyword extraction mechanism are twofold: ??? Feedback from our previous study (described below) showed that the keyword extraction was the least sat- isfying component of our automatic metadata extrac- tion mechanisms to the users. ??? Our data indicated that human annotators often as- signed keywords to a document that do not appear in the document, which were impossible for the current keyword extraction mechanism to generate. Building upon TextRank by Mihalcea and Tarau [4], our ap- proach is to use a graph-based algorithm to rank keywords, based on semantic relationship