A new notion, that of semi-lossless text compression, is introduced, and its applicability in various settings is investigated. First results suggest that it might be hard to exploit the additional redundancy of English texts, but the new methods could be useful in applications where the correct spelling is not important, such as in short emails, and the new notion raises some interesting research problems in several different areas of Computer Science.
Yair Kaufman, Shmuel T. Klein