The usage and purpose of fundamental components like translation memory are being called into question by the emergence of natural language processing, which is already upending the language services sector.
Do we support and encourage this transformation, or do we fight it? Can we work together as a sector, or are we each looking out for our interests?
Let’s discuss the question!
Historical Importance of Translation Memory
Since its extensive use in the 1990s, translation memory has been a pillar of the technological infrastructure of the translation industry. Working with translation memory was groundbreaking at the time. A sentence just needs one translation, and it’s saved for all time. New content could draw on earlier. The cost savings were enormous, and maintaining uniformity quickly became a practical idea, even when dealing with many translators on a single project. Regulatory translation services benefited tremendously from translation memory tools.
Even though it may seem elementary to us now, this was a game-changer. The fundamental ideas of the translation industry has drove for almost 30 years by translation memory. It sets productivity standards and deadlines. Instead of starting from scratch, it gives translators options to either confirm or change. Professional translators working for regulatory compliance translation services have benefited greatly from translation memory tools by decreasing their turnaround time.
Allowing machine translation and translation memories to run simultaneously was the initial move in the changeover. Machine translation will never take precedence over a translation memory feed. It would use translation memory rather than machine translation only if a sentence is stored and there was a 50% similarity between the sentences.
However, as ATNMT has improved, it is no longer a given that a partial translation memory match will outperform an ATNMT feed. A stage is reached where a 50% to 74% translation memory match will require more editing time for a translator to resolve than an ATNMT engine feed for that exact string, according to data gathered from various translation projects. And according to research, we are rapidly approaching a time when ATNMT feeds will outperform 94% of translation memory matches. Professional Biotechnology Translation services have significantly utilized ATNMT.
The existing tools can provide better feeds. Translation memory is not up to the mark. The industry, however, suffers. It is still trying to figure out how to make a single paradigm out of translation memory and ATNMT. As an illustration, Company X might quote while leveraging ATNMT yet pay vendors using translation memory. The majority of existing conventional computer-aided translation systems still rely on translation memory as their main component and can connect to ATNMT engines concurrently.
We must completely abandon translation memory and switch to a more advanced ATNMT architecture. Both payables and receivables should rely on this post-edit effort/distance rather than specific translation memory matching because this model should be able to estimate the required work to finish any given string. And that is something the best biotechnology translation services have integrated into their workflows.
Most translators stated that when translation memory became standard commercial practice, it reduced their ability to write creatively. Moreover, it gave them terrible feeds that took longer to rectify than writing, encouraging human stupidity by locking 100% matches and other adverse effects. A lot of the reaction was justified and relied on preconceptions.
New technology serves its purpose in a different way. Users will have to upgrade their capacity to figure out its utility. The technology is made to address high-level business problems, leaving us simple mortals who use it daily to figure out the specifics. Naturally, it’s likely to be the same with ATNMT. A reality sits somewhere between the idea and the execution.
All involved community members, such as business owners, client stakeholders, and translators, must actively engage in conversation. They must share the same viewpoints.
Managers are concerned with the Return on investment and Profitability. We ignore the language barrier, the translators’ background, and the long-term effects of how innovation affects knowledge management during this process.
People started experimenting more frequently with the idea of post-editing machine translation about ten years ago. They should have considered what this meant in terms of the experience, but it meant more money and production. Imagine swiftly reading through an inconsistent document, occasionally brilliantly written and sometimes filled with unforgivable errors. The compensation is a small portion of the usual cost. And you have to correct any “mistranslations”.
New technology has driven down the food chain without the essential checks and balances. due to high pressure, unclear expectations, and a lack of language and thought leadership from those executing the work. This hasty response to profit generates a miserable experience for individuals affected by the shift.
We are more likely to experience early benefits from this change than fear and frustration if we start by assimilating and then have open discussions about this to establish common ground before choosing rules that make sense for everyone or at least consider multiple points of view.