2004 | OriginalPaper | Buchkapitel
Extracting Citation Metadata from Online Publication Lists Using BLAST
verfasst von : I-Ane Huang, Jan-Ming Ho, Hung-Yu Kao, Wen-Chang Lin
Erschienen in: Advances in Knowledge Discovery and Data Mining
Verlag: Springer Berlin Heidelberg
Enthalten in: Professional Book Archive
Aktivieren Sie unsere intelligente Suche um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Scientific research reports usually contain a list of citations on previous related works. Therefore an automatic citation tool is an essential component of a digital library of scientific literatures. Due to variations in formats, 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 metadata, e.g., authors, title, journal, etc., of a citation string, we present a new methodology based on protein sequence alignment tool. We also develop a template generating system to transform known semi-structured citation strings into protein sequences. These protein sequences are then saved as templates in a database. A new semi-structured citation string is also translated it into a protein sequence. We then use BLAST (Basic Local Alignment Search Tool), a sequence alignment tool, to match for the most similar template to the new protein sequence from the template database previously constructed. We then parse metadata of the citation string according to the template. In our experiment, 2,500 templates are generated by our template generating system. By parsing all of these 2,500 citations using our parsing system, we obtain 89% precision rate. However, using the same template database to train ParaTools, 79% precision rate is obtained. Note that the original ParaTools using its default template database, which contains about 400 templates, only obtains 30% precision rate.