From dda08f40e9abd9a6c6c34072b28865ae0bc1b714 Mon Sep 17 00:00:00 2001
From: Malin Freeborn <malinfreeborn@posteo.net>
Date: Mon, 12 Sep 2022 23:01:54 +0200
Subject: [PATCH] add example pdf-to-text script

---
 data/pdf-to-txt.md |  1 +
 data/pdf-to-txt.sh | 25 +++++++++++++++++++++++++
 2 files changed, 26 insertions(+)
 create mode 100755 data/pdf-to-txt.sh

diff --git a/data/pdf-to-txt.md b/data/pdf-to-txt.md
index 40ae28c..92de8b8 100644
--- a/data/pdf-to-txt.md
+++ b/data/pdf-to-txt.md
@@ -18,3 +18,4 @@ Arch: tesseract-data-eng and poppler-utils
 > tesseract -l eng  "$x" - >> *out*.txt
 > done
 
+- [Example script](data/pdf-to-txt.sh)
diff --git a/data/pdf-to-txt.sh b/data/pdf-to-txt.sh
new file mode 100755
index 0000000..c6da2a9
--- /dev/null
+++ b/data/pdf-to-txt.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+pdftoppm -png input.pdf page
+
+for x in *png; do
+	tesseract -l eng  "$x" - >> out.tex
+done
+
+rm *png
+
+sed -i -ze :a -e 's/\([a-z]\)\(-\)\n\+\([a-zA-Z]\)/\1\3/g' out.tex
+sed -i -ze :a -e 's/\([a-z]\)\n\+\([a-zA-Z]\)/\1 \2/g' out.tex
+sed -i -ze :a -e 's/\([A-Z]\){3}\+\n/\1 XYZ/g' out.tex
+sed -i -ze :a -e 's/\n\([A-Z]\{3\}\+\)\n/\\section{\1}\n/g' out.tex
+sed -i -ze :a -e 's/\([a-z]\)\. \([A-Z]\)/\1\.\n\2/g' out.tex
+
+sed -i 's/“//g' out.tex
+sed -i "s/”/''/g" out.tex
+sed -i "s/’/'/g" out.tex
+sed -i "s/‘/'/g" out.tex
+sed -i "s/\.''/''\./g" out.tex
+sed -i "s/ — / -- /g" out.tex
+sed -i 's/\$/\\$/g' out.tex
+sed -i 's/%/\\%/g' out.tex
+sed -i 's/&/\\&/g' out.tex