1. Advances in Handwritten Keyword Indexing and Search Technologies
- Author
-
Vidal, Enrique, Busch, Hannah, Fischer, Franz, and Sahle, Patrick
- Subjects
ddc:020 ,ddc:004 - Abstract
Many extensive manuscript collections are available in archives and libraries all over the world, but their textual contents remain practically inaccessible, buried under thousands of terabytes worth of high-resolution images. If perfect or sufficiently accurate text-image transcripts were available, textual content could be indexed directly for plaintext access using conventional information retrieval systems. But the results of fully automated transcriptions generally lack the level of accuracy needed for reliable text indexing and search purposes. Additionally, manual or even computer-assited transcription is entierely unsustainable when dealing with the extensive image collections typically considered for indexing. This paper explains how accurate indexing and search commands can be implemented directly on the digital images themselves without the need to explicitly resort to image transcripts. Results obtained using the proposed techniques on several relevant historical data sets are presented, clearly supporting the considerable potential of these technologies.
- Published
- 2017