I agree. My current code inserts email bodies to a compressed table. And that's it. Another simple idea I had was to break up the body by word boundary. Storing it in an array, then doing a bulk insert ignore into a unique column. Might look something like this.... although this a probably a pipe dream. But seems logical... at least based on my stunted repetitive vocabulary. Would have the benefit of being fast due to the lack of compression.

CREATE TABLE words (
    rowid INT UNSIGNED NOT NULL AUTO_INCREMENT,
    word VARCHAR(255) NOT NULL UNIQUE
) ENGINE=InnoDB CHARACTER SET=utf8;
[download]

CREATE TABLE body (
    rowid INT UNSIGNED NOT NULL AUTO_INCREMENT,
    word_order_num INT UNSIGNED NOT NULL,
    word_rowid FOREIGN KEY REFERENCES words(rowid) NOT NULL
) ENGINE=InnoDB CHARACTER SET=utf8;
[download]