“Man vs. machine” becomes “man and machine”
Human-technology collaboration is not only the future, but also the present reality, especially in our field. Machine translation (MT) and artificial intelligence (AI) systems are being integrated daily into the work of linguists, and a symbiotic relationship with the emerging technology is hence more critical than ever. To thrive and survive in this evolving landscape, linguists need to be adaptable, malleable, and willing to reinvent themselves. The advent of these systems will not make human linguists jobless, but will rather reimagine their roles and expand possibilities. Human-tech collaborations can help make knowledge accessible for everybody in a cost-effective, efficient fashion.
To best reap the benefits of human-tech collaborations, it is essential to acknowledge the weaknesses and strengths of both MT engines and human translators. For example, MT plugins can translate just about any (government) website of any size in seconds, but the accuracy and reliability of their output are not yet up to par. This is where the human finesse can step in to refine their accuracy, reliability, acceptability, and cultural sensitivity (where applicable). Language access expansion could start with investing in MT post editing (MTPE). In other words, instead of brushing MT engines or linguists aside altogether, they can work together and complement each other.
Speaking of complementary relationships, two heads think better than one, and while human intelligence is irreplaceable, it is augmentable by AI, especially generative AI. These state-of-the-art systems are the epitome of “you are what you eat.” Therefore, agencies and entities with LEP-oriented content to translate can invest in training and fine-tuning AI models specifically for translating their content. Ultimately, this would help automate and expand language access on the long run, all the while providing LEP persons with high-quality, accurate, and natural-sounding translations before too long.
Term management means bigger impact
Good translation maintains consistency across the board, and there is no doing that without managing content at a granular level, starting with terminology. Agencies and entities with content to translate may invest in creating a public terminology database- termbase- comprising key terms, which creates yet another avenue for human-tech collaborations. AI systems, for example, can help extract terms from a text of any size in just seconds, though arguably — just like the good-old term extraction tools — they produce much “noise” (i.e., unwanted terms) in the process. Linguists can then clean up the noise and document the terms in a user-friendly termbase.
There is a wealth of research on the high return on investment (ROI) termbases can generate; and in this case, the ROI is twofold. A termbase would expand LEP persons’ access to current and future content. It could also help linguists standardize the translation of frequently used terms into other languages, especially diglossic ones such as Arabic. The semitic language has at least three versions (classical, modern standard (MSA), and colloquial) and at least 22 dialects. The diglossic nature of the language has resulted in a certain degree of inconsistency in translating key terms, which has painted terms with a wide brush of polysemy and led to mistranslations in the past.