December 8, 2017

Efforts to gradually improve 9CHRIS have been ongoing, including hand-correcting the identification of documents inside each volume.

Initial correction efforts concentrated on volumes that appeared to be troublesome when analyzed by the document identification algorithms. (This could be because of a large number of drawings or photographs, for example, or very pale title pages.)

More recently we have moved toward hand-correcting every volume, beginning at v0001. We are pleased to announce that volumes 1-1000 have now been done. As mis-identified pages are fixed, they are removed from the 9CHRIS dataset, so search results will also gradually improve. In other cases, documents that were skipped were successfully identified and incorporated into the dataset.

Once all the documents have been hand-corrected, we will have a solid base from which to generate additional linkages and metadata, to improve the ability to find documents of interest.

Please send any suggestions to: