Challenges of Translation: Improving Machine Translation Post-Editing

Challenges of Translation: Improving Machine Translation Post-Editing

In today’s computerized society, a certified translation company in Singapore may find itself using more machines and computer programs to streamline the translation process, and its staff (or freelance translators brought in from outside) may be entrusted with the task of Machine Translation Post-Editing (MTPE), ensuring that the documents translated by an automated process are free from errors and are legible to a regular human reader. In a sense, this is similar to the process of proofreading, the simple difference being that one is now doing it for the final drafts of a computer program instead of a human translator in Singapore. Following that logic, one can follow the usual aims of proof-reading, such as ensuring accuracy and improving legibility. However, the results of a machine can also be radically and sometimes nonsensically different, and that would require you to modify your own thinking and approach to improve your capacity to perform MTPE.

One important thing to realize when interacting with a machine or computer program is that a machine itself is not intelligent. A machine (at least for now) is not sentient, cannot operate on its own, and will never have the initiative to seek out information outside of that which it is explicitly instructed to use. During a PR application translation, a machine may translate the names of peoples and places literally, if it is not instructed to recognize names. This can be a major issue with Chinese Hanzi or Japanese Kanji. In such a case, a certified translation company in Singapore would do well to pay particular attention to parts of the document that chiefly feature names.

Another thing to know is that outside of wear and tear of its components, a machine does not deviate. This may sound perfect for creating error-free documents but remember that machines themselves are not intelligent. They can recognize how individual words are spelled and have instantaneous access to a massive database of words but programming them to understand context is a more difficult process. Similarly, if the machine only understands the spelling of words but not how to use them, it can confuse two different words with similar spellings (one easy example being desert and dessert). When providing a legal translation service, it can involve lengthy documents with crucial words repeated throughout it. A human translator may make a mistake once (“the desert shop named ‘Gobi Dessert Goodies’”) and use the words correctly after that, but a machine that cannot recognize its own error will repeat it throughout the entire document. Fortunately, if you can recognize how the machine is consistently misspelling a particular word, you can use a search function to find every instance of that error and correct it easily.

Ultimately, machines may have their flaws, but so do humans. In our case, we compensate for our capacity for human error with the reviews of multiple people proof-reading human outputs to improve their quality. With MTPE, we do the same for machines. In both cases, this is done to maintain the public’s trust in the certified document translation services provided in Singapore, so that they can be sure that whatever their needs (in particular the typically irreversible notarized translation), they can be confident of receiving a high-quality completely-correct document at the end of it all.

You might like

About the Author: admin

Leave a Reply

Your email address will not be published.