MTPe (Machine Translation with Post-editing)Where time- and budget-constrained translation projects resort to.
More often than not, human translation is found inefficient for some translation projects that require extremely fast turnaround and the budget under control. Some low-budget projects may require output quality differentiation by significance of documents. Machine translation and post-editing could be the most viable solution in such cases.
We never talk about raw machine translation, however. CREDENTIAL knows machine translation alone never helps a translation company come up with a good solution for its clients.
Not being a one-size-fits-all means of translation, machine translation requires good amount of pre-processing and post-editing tasks to produce an ‘output that works.’
This also means machine translation (MT) becomes a viable solution only when these additional tasks cost less than what the equivalent human translation (HT) solution takes.
Knowing how to get the best translation quality out of a machine translation project that involves one or more Asian languages, CREDENTIAL advises our potential machine translation buyers that:
- The source texts of a machine translation project should be modified into the controlled source language
- The source texts have been created in an MT-friendly manner
- In-domain machine translation / human translation memories should be provided to CREDENTIAL
to improve overall machine translation output.
Unfortunately, not all documents suit machine translation.
Among other preconditions for quality machine translation output and MTPe project feasibility, some of the most deciding factors are as follows:
Bilingual Corpora :
Sufficient amount of bilingual corpora to feed the example-based machine translation (EBMT) engine, dedicated to the client for the client-specific domain, help produce quality machine translation output.
Segment Composition :
If the dominant unit in the source texts is words and phrases, such as UI/UA, machine translation works much better compared to the long-sentence segments where literal translation approach is not suitable.
Project Volume :
Use CREDENTIAL’s machine translation service only when the volume of the Project is high enough to make training of engines beneficial (i.e. episodically occurring small translation projects are not suitable for MTPe).