Extracting Citation Metadata from Online Publication Lists Using BLAST

Scientific research reports require a great deal of citation, therefore an automatic citation tool would be of great use. Due to numerous models, it is difficult to automatically transform semi-structured citation data into structured citations. Some digital library institutes, like ResearchIndex (CiteSeer) or OpCit, have attempted automatic citation parsing. In order to recognize citation metadata, we use gene sequence alignment tool to recognize citation data in our method. Known semi-structured citation data are transformed into protein sequences, and saved in a template database. To parse new semi-structured citation data, we can translate it into a protein sequence. We then use BLAST (Basic Local Alignment Search Tool), a sequence alignment tool, to find the most similar template to the protein sequence from the template database previously constructed. We can then parse metadata according to the template. Using the 2,500 templates generated by our template generating system as the template database and parsing all of the 2,500 citation using our parsing system, we obtain 89% precision rate. Using the same template database, ParaTools obtains 79% precision rate. ParaTools contains about 400 templates in the system. Using the default template database, ParaTools only obtains 30% precision rate.