for back-office staff

Split a scanned packet into documents

Split a stack of scanned pages into documents. choose says what kind each page is. decide marks where a new document starts.

Three invoice pages and one notice.

question="What kind of page is this?"
kinds=(
  invoice
  notice
)

cat <<'EOF' |
Invoice 7 for Northwind. Page 1 of 2.
Invoice 7 for Northwind. Page 2 of 2. Total due $1,200.
Invoice 8 for Contoso. Page 1 of 1. Total due $310.
Notice to all customers. Page 1 of 1. New terms from May 1.
EOF
thinkthen choose "$question" "${kinds[@]}" \
  --batch 1 \
  --lines \
  --raw

Output

invoice
invoice
invoice
notice

exit 0: the run finished

awk pairs each page with the one before it. The two pages of Invoice 7 stay together. A new document starts at Invoice 8 and at the notice.

question="Are the two sides from different documents?"

cat <<'EOF' |
Invoice 7 for Northwind. Page 1 of 2.
Invoice 7 for Northwind. Page 2 of 2. Total due $1,200.
Invoice 8 for Contoso. Page 1 of 1. Total due $310.
Notice to all customers. Page 1 of 1. New terms from May 1.
EOF
awk 'NR > 1 { print last " | " $0 } { last = $0 }' |
thinkthen decide "$question" --batch 1 --lines |
jq .value

Output

false
true
true

exit 0

The functions it uses

choosePick one option from your list.decideAnswer one yes or no question about the evidence.

On GitHub: github.com/botassembly/thinkthen