-
Know your Language Service Provider (LSP). Before signing on with them, ask questions and do a little research on their website to ascertain how and whether they use machine translation (MT) and/or AI. This is essential if the material you are having translated, edited, or proofread contains your personal data or that of your customer(s).…
Quantity vs Quality
In language services, quantity means being able to translate, proofread, or edit large volumes of content. Thus, nowadays, most LSPs use machine translation (MT), machine translation with human post-editing (MTPE), or even machine translation with AI post-editing (MTQE). The limited use of human translators means savings for the customer and more profit for the LSP.
On the other hand, quality means a language service performed by humans that captures the original content’s tone, texture, and intention. It involves taking what someone has written, translating it into another language (or proofreading or editing it), and making it sound even better than the original while maintaining the author’s meaning and intention.
Unfortunately, as advanced as MT and AI are, they can never capture the author’s feelings and nuances. This is what is happening in the language services industry today. With the advent of AI, it is now being used as the “human” editor on machine translations. Using only machines means pure profit for the LSP and significant savings for the customer, but at a considerable cost to quality.
I have worked primarily at the same LSP for thirteen years. When I started with the company in 2012, it marketed itself as “only human translation.” I remember asking if I could use a CAT (Computer-Assisted Translation) tool and being told that using any machine was strictly prohibited. A CAT tool has a translation memory, so as one uses it, it remembers previous translations and helps fill in translation, saving the translator time and effort.
Two years later, I was proofreading English for my boss, the company’s CEO, and noticed a shift in his philosophy. He was writing articles about neural machine translation (NMT) and its advantages. That was a surprise. It wasn’t long before some of our translation projects were only available online using a CAT tool the company provided. All the translations done by freelancers using this method went into a vast translation memory within the company’s system.
In 2021, the company was acquired and changed its name, and many changes occurred. Some were good, but most made it more difficult for translators to earn enough to sustain themselves as they had before. Most translations were MTPE, which paid translators 50% of the human translation rate. The company has now implemented MT/QE, machine translation with AI Quality Estimation, with no human involved. If the customer wants a human to check it, then a translator is asked to check it at 1/10th the rate of human translation.
This made me think about language service providers and their policies for inputting legal and/or financial documents into a machine. The machine stores all the information in its “memory,” then it is run through AI, which also stores and retains all the information. What are their policies concerning sensitive personal and confidential information and machine translation? Does the customer opt in, or does the LSP unilaterally decide what to use MT/AI for?
I have asked these questions and more and, as yet, have never received a definitive answer. So, in December 2024, after earning my ATA certification, I started my language service.
My mission is to provide guaranteed, secure, human translations by translators who are certified in their language pairs.
At QA Language Services, customers’ documents are never stored; they are deleted immediately after the translation is completed and accepted. In this way, we safeguard our customers’ privacy and, in turn, their customers’ privacy.
QA Language Services 