Haystack: Docx Support

[anirbansaha96]

I am using Python 3.6.9 and was facing an issue with PDF Converters using PDFToTextConverter and I got it resolved using Issue #182 , and it is now working fine, thank you for the speedy support.

I wanted to know whether there is similar support for Docx files, for now, I need to do this:

from docx2python import docx2python
doc=docx2python('article_text.docx')
text=[]
for sent in doc.body[0][0][0]:
  text.append(sent)
article_text=''.join(text)
import re  
article_text = re.sub(r'\[[0-9]*\]', ' ', article_text)  
article_text = re.sub(r'\s+', ' ', article_text)  

Is there a more direct option like PDFToTextConverter.

[tholor]

It's not implemented yet in Haystack, but would definitely be valuable.
A basic implementation should be rather straightforward. It should mimic the PDFToTextConverter and inherit from BaseConverter. We could use the same library you mentioned (docx2python) to read the file - just not sure if separation into pages is possible there.

Would you be interested in implementing this and creating a pull request @anirbansaha96 ? We could support you along the way.

[anirbansaha96]

Yes @tholor, I'll definitely look into this and try to implement this.

[Utomo88]

Word and PDF documents are the most common documents most people use. It will be great if it is supported.
for eBooks EPUB also common beside AZW and Mobi

[anirbansaha96]

@tholor I have added the functionality and pulled a request #225. Please do look into it.

[anirbansaha96]

Thank You for approving the Merge Request. I've checked in my own implementation. It works fine.