Transliteration
Transliteration is a systematic way to represent the sounds of words in one language using the writing system of another language. The word transliteration is often used interchangeably with transcription. The former is used in modern publications while the latter was used in older publications. Properly used the word transcription refers to the conversion of the spoken word into the written language as in the transcription of the proceedings of a court hearing. For example, Pinyin is a transliteration system for Mandarin using the Roman alphabet, and is often called a romanization. The same words are likely to be transliterated differently under different systems, for example Peking vs. Beijing. Transliteration can be done in a non-alphabetic language too. For example, in a Beijing Newspaper, president Bush's name is transliterated into two Chinese characters that sounds like "Bu4 Shu1" (布殊) by using the characters that mean cloth and weird.
Transliteration has proven to fail miserably in conveying the original pronunciation. One ancient example is the Sanskrit word Channa which transliterated into the Chinese word Ch'an through buddhist scriptures. Ch'an (禪 zen buddhaism) was translaterated from Japanese to Zen in English. Channa to Zen is quite a change.
The idea of transliteration is complicated by the genuine use in multiple languages of different common nouns for the same person, place or thing. Thus, Muhammad is in common use now in English and Mohammed is less popular, though there are excellent reasons for each transliteration. Muslim and Mohammedan are less interchangeable (are they in 2002?), but the typical French usage "Musulman" is considered offensively colonialist in English language contexts.
Another complex problem is the adoption of loan words from one language to another, followed by subsequent changes in 'preferred' transliteration. For instance, the word describing a philosophy or religion in China was popularized in English as Tao and given the termination -ism to produce an English word Taoism. That transliteration reflects the Wade-Giles system. More recent pinyin transliterations produce Dao and Daoism. (See also Daoism versus Taoism.)
In the study of languages written in cuneiform, transliteration referrs to 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 polyphony, in which more than one sign represents a given sound (look polyphony and polyvalence up!), 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". Due to a historical artifact, the sign number one 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".