Abstract: Currently, in our digital library, dissertations database is one important of digital resources. However, metadata entry has relied on manual to complete, which is low efficiency, and cost a lot of manpower. For this problem, our applied the method of document features and pattern matching, and made use of regular expressions to research automatic extraction of dissertation metadata. The algorithm includes two modules of information field location and metadata extraction. The experimental data shows that the algorithm has higher precision and recall, and overall performance index F.

Key words: Dissertation