In steganography, the plaintext is embedded without affecting the meaning of the medium or the plaintext. That is why the covertext provides no clue of embedded secret in it. Recently, a dictionary-based text steganography is proposed that utilizes a dictionary of meaningful English sentences to be used as covertext for a given plaintext. Here, the candidate covertext is selected at random. In this paper, we further extended the work by proposing a new way to select the candidate covertext. The proposed techniques is able achieve uniqueness in stegokey when repetition of characters are involved, while achieving better covertext to plaintext ratio and decoding time.