nlp-private:noisy-ocr-group

__TOC__

Planned Meeting Topics

For Now:

Review annotation tools, e.g. Lehigh's http://dae.cse.lehigh.edu/DAE/, GEDI, PixLabeler.
Dan and Dr. Ringger present a good paper on relevant current research.
Get CIKM “AND” proceedings, '07, '08, '09. Very relevant venue for this group.
Josh Hansen's HMM-LDA project?
Report on state of data.
Decide on an annotation tool (Image Annotation Tools)
Decide on an OCR engine (OCR Engines)
Document image data sets (Document Image Data Sets)
Funding discussion.
Aaron to demo FOCIH-based image annotation tool.

For Later:

Create annotation plan for Ancestry and other data.
Hire under-grads, library, request Ancestry employees to do annotation.
Dr. Ringger to invite Dan Lopresti to come to BYU.
Decide to make a competition.
More clarity on directions and contributions, including specific projects, what tools to use (Ocropus?), how to leverage those tools, what data to work on, how to magnify our efforts by working together as a group, etc.
Plan for zoning, layout, language modeling, table interpretation and E/R classification earlier in the (Ocropus) pipeline.
Should we use Ocropus, train a character recognition model from Internet Archive data, hire an undergrad to do this plus output bounding boxes, etc.?
CHURP to NSF Proposal (sometime in June to August).
OCRopus update.

These are big tasks affecting everyone in the NOCR group.

Image Annotation Tools: Find, adapt or create an annotation tool to help us create good benchmarking data sets, including gold standard zone, entity and relationship annotations.
Document Image Data Sets: Find, annotate and extract from one or more document image data sets.
OCR Engines: Find a good OCR engine and use it.
Annotation Guidelines and Metrics:
Papers