1. Contextual Ambiguities

One of the primary challenges faced by AI translation systems is the interpretation of context. Languages often rely heavily on contextual cues, idioms, and cultural references that may not have direct equivalents in other languages. AI models, lacking the ability to fully grasp the nuances of context, can misinterpret phrases and produce translations that may seem accurate on the surface but fail to capture the intended meaning.

Example: In 2017, Google Translate made headlines when it translated the Ukrainian phrase “Росія – це моя батьківщина” into Russian as “Russia is my country.” However, the correct translation was “Ukraine is my motherland.” The mistake sparked controversy due to the ongoing political tensions between Russia and Ukraine, emphasizing the importance of accurate translations in sensitive contexts.

2. Cultural Sensitivity and Nuances

Cultural sensitivity is a critical aspect of language translation, and AI systems may struggle to navigate the intricate cultural nuances embedded in languages. What may be acceptable or even humorous in one culture can be offensive or confusing in another. AI models, lacking cultural awareness, may inadvertently produce translations that are inappropriate or convey unintended meanings.

Example: In 2018, Facebook’s translation algorithm faced criticism for translating a post from Burmese, a language spoken in Myanmar. The translation referred to the ethnic minority group, the Rohingya, in a way that was deemed offensive and potentially fueling ethnic tensions. The incident highlighted the need for AI systems to be more attuned to the cultural and political sensitivities of different regions.

3. Polysemy and Homonymy

Languages often contain words with multiple meanings, known as polysemy, or words that sound the same but have different meanings, known as homonymy. AI translation models may struggle to discern the intended meaning of such words, leading to errors in translation.

Example: In 2020, Microsoft’s translation service raised eyebrows when it translated the Chinese phrase “男孩在跑” into English as “The boy is running.” However, the phrase can also mean “The boy is in the bathroom,” showcasing the challenge of accurately translating ambiguous statements without additional context.

Conclusion

While AI translation has undoubtedly transformed global communication, blunders you typically would not see in human translation underscore the complexities and limitations that still exist in the field. As developers continue to refine and improve AI translation models, addressing contextual nuances, cultural sensitivities, and linguistic ambiguities will be paramount. The journey to seamless AI translation involves not only advancing technological capabilities but also fostering a deep understanding of the diverse linguistic and cultural tapestry that makes human communication rich and complex.

The post Artificial Intelligence (AI) Translation Blunders appeared first on Anthony Teixeira - Professional French Translator.

No, I have nothing against MT

In fact, I use it in my translation tools! It’s a great way to generate lists of words for predictive typing, which indeed saves me a bit of time and reduces stress for my fingers. As as I said a few times as well, MT can be great for research purposes, especially for individuals. For these reasons, I am happy to see MT improve, both as a translator and as a human being.

Yes, statistical MT engines are improving

MT engines of today are smarter and keep learning from bilingual texts. It means that compared to legacy systems, they have a much better understanding of context and can string together pretty decent sentences if the right references are available. I have no doubt they will keep on improving, as long as there are people to train them.

No, I’m still not interested in PEMT projects (for publication purposes)

In my previous post, I made the mistake of referring to PEMT projects as if they were all identical. Post-editing machine translation for training purposes is something I would have no problem with. I would charge an hourly fee for this and all would be good.

The projects I mentioned (or meant to mention) are those for which we, translators, are asked to provide a discount to edit texts that went through some MT engine. Usually, this will be Google Translate or a very slightly customized version of it. The goal of people offering such projects is to get a translation ready for publication at a reduced price, which has nothing to do with MT engine training.

Here, we have two possible cases:

-The text was processed through a generic system, say Google/Bing Translate. These engines are using bilingual texts from sources that are so different in terminology, style, quality, etc. that the resulting text is generally very inconsistent and poorly written.  Why one would want to give a discount for a text you need to rewrite from zero?

-The text was processed through an industry-specific, well-trained engine. Let’s go even further and suppose the target text comes together near-perfect. It still doesn’t cut it for me. When all I have is the text generated by the MT provider, I still have to look very carefully at the source and target segments and make sure everything is there and properly translated. Again, it can be very, very easy to let a mistake slip (say, a missing negation) because the MT engine used its “best match” and failed to translate the part it had no reference about.

If the generated translation is good, it means a very close translation already exists somewhere and the engine knows about it. In this case, at translation memory would work better, as it would allow me to see where exactly the new text is different and where I have to be careful. If the TM is approved and no proofreading is required for the existing part, the changes can be done very quickly, much more than if I had to check the whole segment.

To put it simple, a good MT output means there is a good TM available out there (maybe not in a TM format as such, but something that could be converted easily), and I’d rather work with the TM in question for practical purposes.

To sum it up:

– MT is useful

– PEMT for training purposes is fine

– PEMT to produce a professional translation is a waste of translator and client money

The post I have no problem with MT, but PEMT still doesn’t work for me appeared first on Anthony Teixeira - Professional French Translator.

The increasing number of Machine Translation Post-Editing (MTPE) jobs posted online seems to be one of the big trends of the translation industry. These jobs essentially consist in fixing translation provided by an automated tool (Google Translate, Bing Translator, etc.) for a lower price than ordinary, 100% human translation.

The idea sounds good on paper, but faces a major issue:

MTPE takes more time and energy than human translation for poorer results

A translator will typically read the source text, think about the translation and write it out. MTPE adds a step before that: comparing the source to the MT output. It is tempting to think that because MT engines output a draft, translators save time when typing the translation. The truth is that, often rather than not, the output will require so much rework that it would be faster to type the translation out from the start. On occasions, the MT result will need only minor rework, but the time saved here is taken away by the comparison bit I mentioned earlier.

Even if the best of the cases, MT doesn’t save you time. And most of the time, it will require more efforts than a human translation.

On the top of that, the final quality also suffers, and that for two reasons:

– With MTPE, editors fix a text to make it “acceptable”, readable. Human translators try to produce texts that are fluent, natural in their native language. Something “good” rather than merely “comprehensible”.

– Because of the way MT engines work, the output can occasionally contain very serious mistranslations. Google Translate will very often omit words, especially negations, because it looks for similar sentence patterns in its translation memory rather than translating from scratch. Of course, the job of a MT post-editor includes spotting these mistakes, but it’s easy to let one slip away when you are correcting hundreds of them for hours.

MTPE doesn’t make sense for end clients either

– For a high quality translation, asking a professional translator is the best way. Turning MT output into a good text will end up costing more than what a good translator will charge

– If you need a text translated for information purposes only, MT does that very well already. Of course, there will be grammar mistakes and even occasional mistranslations, but in most cases you will be able to understand the general idea behind the document.

MTPE is a hybrid approach that mixes the disadvantages of both MT and human translation: you will end up paying a price very close to that of human translation, for a result barely better than what Google and Bing give you for free. Half-baked concepts rarely produce good results, and MTPE is no different. I never accept post-editing projects because they don’t benefit anybody down the process, from the translator to the end customer.

The post I Don’t Offer Machine Translation Post-Editing Services, Here Is Why appeared first on Anthony Teixeira - Professional French Translator.