Machine translation (MT) systems are unlike all of the other tools described in this glossary, because rather than assisting a human translator or other language professional in his or her work, MT systems take charge of the entire process of translating texts. However, this doesn't mean that language professionals have no role to play. When MT systems are used, humans are most often involved in the revision (called post-editing) of the target text produced by the MT system to ensure that it is correct and adequate for its intended use. In some cases, humans may also be able to adjust the way the MT system works (e.g. by adding to or modifying the dictionaries it uses) or by preparing documents in such a way that they can be translated as successfully as possible by the system (called pre-editing). MT systems are most useful when source texts can be carefully prepared to be easily translated by the system (e.g. by clarifying ambiguous expressions, using short sentences), and/or when the target text is intended purely to assist in comprehension (and not, for example, for publication). There are different underlying techniques used by MT systems. Some try to imitate the ways in which humans process language (e.g. using grammar rules), while others operate using statistical probabilities or by taking examples of previously translated text as models. Just as human translators may produce slightly different versions of a target text, so will different MT systems. It is important not to mix up the short form MT for machine translation with TM, which stands for translation memory. You will find tools such as Systran and Reverso installed in the Writing Centre; a number of other systems, such as Babelfish and Google Translate, are also available online for free. (It is important to be particularly cautious when using free online tools, as they obviously do not offer the same kinds of advantages as locally installed tools that can be adapted for use on specific texts!)