In the study of languages written in cuneiform, transliteration is the process of representing the sounds of written cuneiform signs in a lossless way, as opposed to transcription, which is a lossy method of representing the spoken language. Because cuneiform is polyvalent , signs may be interpreted to represent more than one syllable (or logogram). For example, the sign DINGIR may represent either the sound "an" or "il", as well as the word meaning god and the phonetic complement for a name of a deity. Similarly, the sign "MU" represents either the sound "a" or the word meaning water.

Therefore, a text containing DINGIR and MU in succession could be construed to represent the words "ana", "ila", god + "a" (the accusative ending), god + water, or a divine name "A" or Water. Someone transcribing the signs would make the decision how the signs should be read and assemble the signs as "ana", "ila", "Ila" ('god"+accusative case), etc. A transliteration of these signs, however, would separate the signs with dashes "il-a", "an-a", "DINGIR-a". This is much easier to read than the original cuneiform, but now the reader is able to trace the sounds back to the original signs and determine if the correct decision was made on how to read them.

Since cuneiform also exhibits polyvalence , in which more than one sign represents a given sound, the transliteration for a phonetic value includes a designation of which sign represents the sound. Cuneiform signs are canonically numbered, and usually a subscripted number follows each sign: "u₆" corresponds to a specific sign, whereas "u₄" corresponds to a different one, both of which are pronounced "u". By convention, the first sign is unnumbered and unaccented: "u" = "u₁", number two is unnumbered with an acute accent: "ú" == "u₂", and number three is often unnumbered with a grave accent: "ù" == "u₃".