a basic starting point to get IA styled content in djvu form, break them up into pages to get them ready for re-OCR'ing them. you need djvu tools (djvulibre-bin in debian/ubuntu)

oicr.sh

#!/bin/bash

[[ -n "$1" ]] || { echo "Usage: oicr.sh IA_BOOK_TITLE"; exit 0 ; }

# sample record ids: catalogueoflepid02briti electronicnaviga00unit halfhoursinfarno00newy nachrichtsblattd3234190012deut

BOOK=$1
BASEURL=http://cluster.biodiversitylibrary.org/n
#BASEURL=http://www.archive.org/download

mkdir -p  $BOOK/ocr $BOOK/xml $BOOK/djvu $BOOK/src

if [ -f $1.djvu ]; then
	echo "$1.djvu found, contining without download..."
else
	wget --user-agent="Mozilla/4.0 (compatible; MSIE 8.0; Windows NT 6.2; Trident/4.0; SLCC2; .NET CLR 2.0.50727; .NET CLR 3.5.30729; .NET CLR 3.0.30729; Media Center PC 6.0)" --tries=20 --span-hosts --recursive --level=1 --continue --no-parent --no-host-directories --reject html --cut-dirs=2 --execute robots=off $BASEURL/$1/$1.djvu
fi

cp $1.djvu $BOOK/src
djvmcvt -i $BOOK/src/$1.djvu $BOOK/djvu index.djvu

ls -1 $BOOK/djvu/*.djvu | grep -v index.djvu | cut -d'_' -f2 | cut -d'.' -f1 | while read PAGE_NUM
do
	echo -n "writing $1_$PAGE_NUM"; echo -n "	ocr..." 
	djvused $BOOK/djvu/$1_$PAGE_NUM.djvu -e 'output-all' > $BOOK/ocr/$1_$PAGE_NUM.txt
	echo -n "done"; echo -n "		xml..." 
	djvutoxml $BOOK/djvu/$1_$PAGE_NUM.djvu > $BOOK/xml/$1_$PAGE_NUM.xml
done

echo "done"
exit 0