InkCapture is software that profoundly expands the ways in which historical handwritten documents can be digitized for archival and information retrieval purposes.
The technology for digitizing printed documents, whether new or ancient and rare, such as books, newspapers, and even maps has become commonplace; such technology is essential for rapid and direct searches of text in documents, as well as for the efficient and cost-effective means for the storage and archiving of precious documents.
But until recently, there has been no such technology specifically suited to handwritten texts, such as chronicles, diaries, notes, handwritten documents, and other antiquities. Lacking reliable tools for automatic handwriting recognition, researchers have had to devote many hours to scour through hundreds of pages of documents without any supporting text search capabilities.
InkCapture aims to change this. By employing the latest developments in machine learning and computer vision, it will enable the automatic retrieval of text information from handwritten documents and so exponentially increase the speed, efficiency, and accuracy of searches in historical handwritten documents.
Handwritten text recognition using neural networks
InkCapture uses machine learning methods and neural networks, which allow the system to recognize fonts written in different styles. InkCapture needs as many historical manuscript documents as possible in the learning phase, since it has the ability to learn and build its own recognition memory, and its success is greatly influenced by the amount of data on which it can train letter and word recognition.
Digitization of manuscripts using artificial intelligence
Municipal chronicles, archival documents and documents, private historical documents, hand-kept records