The advance of machine translation (MT) and its implications for the translation field were prominent themes discussed at the ATA conference this year, so Audra Lincoln’s presentation topic on effectively editing Japanese texts for use with these tools was well-timed! Audra (who goes by they/them pronouns) offered tips and advice to get a better translation using these tools based on their own experience as an automotive translator.
Audra explained that while MT can be an effective tool, it often produces a poor output that requires lots of post-editing. They explained how pre-editing your text can help MTs avoid errors, requiring less editing afterward. Audra classified the issues translators may encounter feeding Japanese through MT programs as either Style or Language problems. They showed how different MT applications (Google, Windows, and DeepL) translated the same text, then compared the accuracy of the results.
Style issues are things like the direction of the text, spaces, hard returns, symbols, or using katakana. Audra gave the example of 工程 さかのぼりinput vertically. The MTs could only decipher it horizontally. They also explained how MTs struggle with spaces between words, hard returns (line breaks), and symbols. The machine treats the text that comes after these as different sentences completely. Katakana can also be an issue because MTs often leave it as-is (like ナゼナゼ分析). Simplifying the format of your text can solve these errors.
Language issues include shortened words, grammatical number, perspective, and polysemy. Audra suggested caution using shortened words like トルコンand チョコ停, which are difficult for MTs to translate. MTs also struggle with quantities and numbers, since plurals are not always clear in Japanese. For perspective issues, MTs will not know the “who, what, when and why” of the text unless specified. This goes along with Audra’s caution about words with multiple meanings (polysemy). You may need to add more context when prepping your text for machine translation.
The key takeaway from Audra’s presentation was to “know your MT tool,” especially its strengths and weaknesses, as some will handle Japanese documents better than others. As a fellow automotive translator, I certainly appreciated Audra’s examples taken from their own work experiences and will use their advice in my own career!
■Lauren Casey
Leave a Reply