Searching Digital Political Cartoons

The study of cartoons, manga, and graphic novels is of growing importance to humanity scholars. Managing cartoons for scholarly use presents two challenges: searching and understanding. Current cartoon archiving and searching projects either organize cartoons for browsing or manually record the symbols and the text in cartoons for searching. We propose to automatically detect the text lines in the cartoons, recognize the text using OCR, expand the scant text evidence in cartoons using related external text (such as news) to facilitate searching, process the text into character trigrams, and use the related external text as a context to facilitate the understanding of the cartoons. Preliminary experiment results are reported.