Add E-Mails to training data by IMAP Server or MBOX, EML, MIME file

I have thousands of emails from interactions with customers, answering questions, that would be a great source of training data for the bot.

Of course not every E-Mail in my archive is useful. I would prefer either adding some common E-Mail file formats like MBOX or EML, MIME or by adding IMAP credentials and then listing all mails on the server in main and subfolders and let me select multiple mails for training. 

What do you think? Any other users who would like to use their company mails for training the bot? I know it could also be done by converting single mails to PDF and then adding these PDF, but this would be a very time consuming, slow way.