While there are sophisticated and efficient steganographic schemes with images as cover available, I am yet ignorant of the existence of any fairly efficient and secure schemes with texts as cover. This state of affairs is IMHO unsatisfying, since the communication channel may under circumstances only be able to transfer printable natural language texts, and that in rather small volume, due to constraints posed by the warden.
To be concrete on the efficiency issue, I would desire that the ratio of stego bits to covertext bits be no less than, say, 1/50. For references to existing schemes, if any, or else novel basic ideas that could lead to the development of such a scheme (if possible user-friendly and not too hard to implement), I should be very grateful.
[Addendum 1, edited] I have in the meantime a scheme named WORDLISTTEXTSTEGANOGRAPHY that should achieve a not large but IMHO practically nonetheless useful efficiency in the range of [0.5, 1.0] bit per word of natural language texts, requiring though that the user does some amount of trial and error work under the guidance of the software. (Use e.g. of an appropriate codebook could highly compress the input stegobit sequence.) The latest Version 2.1 of it, running on Python V.3.6.1, is available at my new home page: http://mok-kong-shen.de. See update notes in it for the differences to the earlier versions. (I regret in particular that in the earlier versions a sentence is missing in the explanation of how to do the examples, with the very unfavourable consequence that new users may fail to do the examples.)