Development of a prototype system for automatic database construction from academic journals

Abstract

A prototype system was developed for automatically constructing a database from scientific and technical journal aicles in print. When a page of a printed technical paper is scanned with an image scanner, the system distinguishes between tables, graphs, and texts. And then in the texts, the logical attributes such as the title, the authors, the author's affiliations, the abstract, are recognized by using the page layout model. Lastly, the texts are processed with an OCR. The performance measurements show that the system has a sufficiently high recognition rate and processing speed and can come into operation.