 |
 |
The crowd sourcing project ran from Thursday, August 26, 2010 to Monday August 30, 2010. Over those 4 days, 53 people evaluated 7,779 images. We want to thank all of you who participated. Below is a table showing how images were categorized during the project:
|
Category |
Number |
Percentage |
|
good |
3396 |
46% |
|
close |
1902 |
26% |
|
bad |
1153 |
16% |
|
disaster |
753 |
10% |
|
don't know |
39 |
1% |
|
irrelevant |
442 |
----- |
Based on closer examination of the images that were judged to be “bad” or a “disaster,” we have 3 areas of focus for future development:
(1) We will use a method from another project to identify large initials, which proved quite problematic for the tool.
(2) We will reduce the number of two column manuscripts in which the left column does not have its lines properly detected by modifying the way the column boundary is drawn, so it does not include the outside border.
(3) We will work to detect marginal content as such, and parse lines of marginal content seperately from the columns of text.
We hope to begin work on our transcription tool in the beginning of October, and release this detection tool as an API by the end of October.
So thank you again for all your hard work.
Jim
This project is being funded an award from the President’s Research Fund, Saint Louis University. The prototype line parser was first developed during a grant from the Andrew W. Mellon Foundation. The Center thanks both funding bodies for their support of this kind of research.
All images displayed are owned by the Matthew Parker Library, Corpus Christi College, Cambridge.
© 2009 Masters of Corpus Christi College.
Images used with permission.
No one may duplicate any image (in digital or printed format; or store on an electronic device) without the express permission of the Masters of Corpus Christi College.