metadata
license: apache-2.0
This dataset was contributed by the creators of the Bayesian spam filter SpamAssassin and can be found at [SA] in various separate files. I think the latest readme total count is off by 1, as I found 6,046 messages, of which 1,896 are labeled as spam. The messages include all headers, although the contributors did apply some address obfuscation and hostname replacements. I aggregated all messages into a single CSV file, including a column with the original file name of each email. As I am focussing on the NLP aspect, I combined the Subject field with the Contents of each email into one column in the CSV file.