Academic journal article CLCWeb: Comparative Literature and Culture

Intermediality and Human vs. Machine Translation

Academic journal article CLCWeb: Comparative Literature and Culture

Intermediality and Human vs. Machine Translation

Article excerpt

In translation studies faithfulness in literary translation exists only to some degree. Since unfaithfulness in literary translation is a matter of definition, the acceptance of relatively faithful but imperfect translation acquires new contexts in digital humanities (see, e.g., Scott; Huang). From an intermedial point of view, a translated text may be considered a new or hybrid product that does not have to be evaluated solely against the primary standards of the source language or its author's culture. Instead, such primary standards may be reduced to secondary in quality assessment. In this article, I address the issue of imperfection in machine translation (MT) versus human translation (HT). Both forms of translation involve a process of the transfer of meaning or knowledge including culture and other elements, and are thus treated as equals.

Since its beginning in the 1950s and 1960s, the use of machine translation includes technical documentation (see, e.g., Hutchins, "Computer-based Translation"). Methodologically, research has gone through the beginning a trial-and-error stage followed by corpus based approaches in the late 1980s. There have been the "direct translation" model and the "interlingua" (indirect) model, including a large number of systems many of which have been used by government departments and corporations. The 1980s then saw the growing interest in spoken language translation. After two decades of research and development backed up by fast-speed computers, MT has been available to many individual internet users. However, what may be described at present is that much of online automatic translation is inaccurate. Nonetheless, one is reminded that since authors, such as the Chinese literary icon Lu Xun (see Huang, "The Translatologese Syndrome"), also experience difficulty in expressing their ideas, and that since translators never produce perfect translations, one has no reason to expect flawless translations from the computer. In the process of transferring meaning in the translation from one language to another, from print to electronic form, leads to a fundamental change in communication (see, e.g., Sager 256-58) resulting in another medium. Moving electronically translated texts to the internet, including the yet unpopular simultaneous speech translation, presents itself as a third medium. All of these intertwine, interline, depending upon each other (see, e.g., Chapple and Kattenbelt; Lopez-Varela Azcarate and Totosy de Zepetnek). One bottleneck problem that remains unresolved is the lack of standardized quality assessment. Although MT evaluation has become an important aspect of research, no formula or easy-to-apply model has been created either for MT or HT quality assessment (see Hutchins, "Machine Translation"). By and large, frontline evaluators assess translated texts on a piece-by-piece basis, while scholars attempt to create models and approaches that measure TT against a non-existent perfect product and unaware of the dividing line between acceptability and unacceptability.

In the present article, the data used in the quantification of the relevant issues come from an international survey where three literary excerpts translated into English from Chinese were surveyed: about 300 professional translators--including 15 senior United Nations translators--completed the different versions or different parts of the international survey (see Huang, A Model for Translation). One question was to find the maximum rate of inaccuracy in HT that can be tolerated by the international community of translators, writers, editors, and translation scholars. This maximum number thus becomes the ceiling under which a TT may not be rejected, but over which a TT is considered a failure. Expressed in numerical terms, this ceiling becomes the dividing line between TT acceptability and unacceptability. Another question was to discover the maximum inaccurate rate in MT which the professionals could tolerate before flatly rejecting it. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.