Automatically translated websites

amk

There seem to be a growing number of websites containing an automatically translated content. Now, there's nothing wrong with that per se, as long as you find what you're looking for, but I'm yet to see well machine-translated article. They can mostly do the title right, some sentences might even make sense, but overall, it's a disaster.

I always block these when I encounter them, so here are some examples:

https://pest.bigbadmole.com/cs/nasekomye/babochki/kak-gusenitsa-prevrashchaetsya-v-babochku/
https://www.actualidadmotor.com/cs/los-frenos-ceramicos/
https://cs.avktarget.com/articles/nauka/v-chem-raznica-mezhdu-massoj-i-vesom.html (the title literally says What's the difference between weight and weight?, perfectly illustrating my point)
https://masterwarm-cs.techexpertolux.com/plastikovye-okna/wds-300.html
https://cs.delachieve.com/bus-male-tridy-paz-32054-historie-a-popis/
https://avtotachki.com/cs/golovnoy-dvigatel/
https://technoluxpro.com/cs/stati/kia-ceed.html
https://adrianysus.com/cs/wat-is-oxidatie-in-wijn-en-kun-je-geoxideerde-wijn-drinken/
https://my.electricianexp.com/cs/ot-chego-podklyuchit-vytyazhku-v-vannoj-komnate.html
https://outdoorfitnessandfun.com/boxing-a-manly-history-sweet-science-bruising

I don't have specific queries, but you can easily find them by just copying keywords into Kagi. These are just random pages found on them. Every single one of those is an automatically and badly translated website, usually from Russian, but other languages such as Spanish or French are no exception.

I can mostly spot them pretty fast just by looking at the URL - I specifically encounter such results with queries in Czech (but I know for a fact that most of these websites are translated to Slovak, Hungarian, Polish and probably other languages as well) and usually don't expect a never seen before domain with .com TLD (this by the way is interesting, given where from these sites originate - it always seems to be a .com) or containing non-czech/non-english words with /cs/ prefix in path.

I recall Google giving me such results as well back when I was using it, so it seems to be a common problem.

Is there any chance of filtering such results out?

carl

Looking at one of your examples, filtering out this tag will help get rid off some of these sites:

<meta name="uri-translation" content="on" />

This tag comes from a Joomla auto-translate plugin.

My suggestion: Identify ways to filter sites with specific meta or code and crowd-source that as a block list.

amk

carl That's a good observation, but it's not a definitive answer to the problem. That meta tag is not meant to indicate machine translated page and it's going to disappear the moment they learn it's used to detect this. Also, there are legitimate sites running Joomla serving properly translated content.

One way of telling this is happening could perhaps be observed from the fact that a site suddenly offers hundreds of new entries in a language that hasn't been there before, although I have no idea how these are born. Maybe that's why they almost exclusively run on the .com TLD, moving content from country specific domains, making this much harder to spot.

MightyPork

I keep getting a lot of these that are usually poorly translated from Russian (maybe indexed because of Yandex?).

Some contain outright incorrect information, maybe by GPT, but mostly it's very bad automatic translation. The problem is that I don't notice this immediately and waste time reading the article before it stops making sense.

Often there are also illustrations with Russian text embedded.

This really degrades the experience of using Kagi.

Here are some examples on my blocklist:

povedet-cz.eu
scoolbylife.com
adinfinitum.cz
desaterotvariosobnosti.cz
kraldilny.cz
stavba.taktojenassvet.cz
florianus.cz
poofi.cz
1domacikutil.cz
2staveniste.cz