PDFBox integration for #30

xpmethod · May 26, 2014 · 97692f1 · 97692f1
1 parent 7e11bda
commit 97692f1
Show file tree

Hide file tree

Showing 3 changed files with 9,424 additions and 0 deletions.
diff --git a/extractor_research/extractors/pdfbox-app-1.8.5.jar b/extractor_research/extractors/pdfbox-app-1.8.5.jar
diff --git a/extractor_research/extractors/pdfbox.py b/extractor_research/extractors/pdfbox.py
@@ -0,0 +1,15 @@
+#!/usr/bin/python
+import os
+
+class PDFBox:
+    def __init__(self, pdf_file, txt_file):
+        self.pdf_file = pdf_file
+        self.txt_file = txt_file
+
+    def extract(self):
+        command = 'java -jar pdfbox-app-1.8.5.jar ExtractText ' + self.pdf_file + ' ' + self.txt_file
+        os.system(command)
+
+if __name__ == '__main__': 
+    pdf = PDFBox('../input/pride_and_prej/1.pdf', '../output/pride_and_prej/pdfbox/1.txt')
+    pdf.extract()