Would it make sense to maintain the AI icon after we chose the definition? So that we remember that that selection was coming from AI and not from the popular meaning? What do you think?
3 Likes
First of all it should be working properly. Besides that, as much as I often question the ai suggestions, the popular ones aren’t much better either.
What would be worth considering, if implementable within a reasonable amount of work, is to not add ai generated meanings to popular ones if the user picks those.
Even better would be to show translations taken from dictionaries, as they are usually way better.
And the best option imho would be to take the definitions from monolingual dictionaries translated for us via ai. Simple because they cover the most ground. My current workflow is: guess proper translation based on suggestion and context->check German translation->English translation->monolingual entry copy-pasted to google translate. Another possibility is to google the word and look for the images that pop up.
However, the ai now even suggests words that doesn’t even exist. And if I translate the whole lesson with the ai set to google translate, it basically ignores every grammatical rule that exist in my mother tongue. Which it doesn’t if I use google translate directly. I am no expert in this matter, but to me it appears as if the issue doesn’t lie in the ai only, but also the way it gets incorporated here.
1 Like
yeah, I don’t really understand what’s the big deal to have an integrated dictionary. I have just used both ReadLang and Language Reactor, and they both have an integrated dictionary with a simple direct translation, plus an explanation, and examples. If they don’t have copyright problems, why LingQ couldn’t figure out the same for all this time! I need to check another software in the next weeks too.
I have to say that with AI definitions I go faster because I quickly check with the popular meanings and opening a dictionary. If they match, I just quickly click it, if I want to read faster.
I just would like to remember that I have chosen an AI option, that could be an interpretation based on context and not the real word definition.
But maybe it’s not so important. That’s why I was asking if it would make sense or not.
That’s what I do, too. However, depending on the text this turns out to be a big if.
I understand that. But popular meanings are often no real word definitions either.
If you mark ai based translations because they could be faulty, then in theory we would need to do this for all definitions, as we cannot judge where something a user typed in came from.
It’s also a matter of language pairing, so which language one learns compared to which language one already knows. If I search for a german translation for a spanish verb, for example, I easely get 20-30 different translations sometimes covering all kind of things that word could potentially mean, based on context and how literally it is meant. Doing so with a Korean word does not provide me with such a huge amount of information. So it is always a guesswork to a certain degree. I often find myself typing a definition that is neither the one found in a dictionary nor in the suggestions made (may it be user or ai-created), but some sort of middle ground that reflects the general meaning of those suggestions and fits the context. My approach has always been: “if the translation makes sense, it is probably right - or at least not complete nonsense”. 
2 Likes
That’s true. The fact is that I always recognize my own definitions from the popular meanings when I click a yellow word. That’s why I thought about to keep the AI icon. In this case I would always recognize my own definitions, the popular meanings, and AI.
But I understand that for the vast majority of users can’t be like that, so it doesn’t make sense.
You may not be marking them because they are faulty, but because they are AI. The whole problem described by OP would be solved if there was some sort of distinction between AI translations and human translations. But I don’t know how they actually work, is there any confirmation that they are always the same in the same context? Are they really rated together with the user definition? I don’t think they should. But in general I don’t think I’ve been afflicted by the issues I’ve seen people reporting. I also don’t want to get much into how others use LingQ, but I think this excessive fine-tuning of translations may be detrimental.
I also think the icon for the AI should be changed, maybe some [AI] mark, had I not seen the forum announcement I might not immediately gather that it was an AI contextual translation.
1 Like
I just want this “feature” entirely disabled, and soon, as it’s really ruining my user experience… I know I’m far from alone.
4 Likes
The issues encountered with ai (nonsensical translation, translation of whole phrases instead of just the word, …) can and, at least for me, have been encountered with user generated translations, too. The main reason imho for reporting this is that, in difference to user generated translations, ai translations can be improved. There isn’t much the LingQ staff can do in regards to what the users create 
It has already been stated that such an option will be added.
1 Like
Im learning Ukrainian. As described, often now when I click a blue word, it doesn’t translate just the word, but the adjacent words as well, forcing me to manually alter the LinQ to remove it. Otherwise, viewing the same word in a different context would show the AI translation which includes words from the original context.
This is really disruptive and I would like to disable this feature.
I believe this feature would remain satisfactory if only the word alone were added to the LinQ.
2 Likes
I’ve noticed this too. It seems to be a new bug, as it didn’t use to happen - or if it did, I didn’t notice it happening before a week or two ago. Fortunately, in the languages I’m learning, I’m advanced enough that I can spot when the translation software is making a mistake, but I fear people new to a language might get badly misled.
3 Likes
Same thing, it didn’t happen before. I think they have tweaked something and it is not working as intended. This is a characteristic that can be adjusted by improving the prompt they have given to the AI translation. I hope we don’t have to wait forever for that to happen.
@Pr0metheus @davideroccato
For me it happened from the very beginning. It may depend on the language or how far one has already proceeded in it, I don’t know.
Did you have it enabled when they were testing it? There was an option to enable it before they released it to everyone. It was only on iOS at first, then they moved it to the web app. I don’t know about Android.
When I say before, I mean that early testing period.
I am using LingQ on Windows 10 only, Chrome browser. I’ve used this feature after it was announced in the forum. I’ve never seen some sort of “beta access” button or similar that would allow access to options that are work in progress.
I’ve noticed, though, that the ai feature overall seems to become worse over time. A while ago I’ve translated a whole lesson via ai and almost every single sentence of the translation generated was grammatically wrong.
The ai sometimes switches between German and English, too, when creating a translation. This happens both when translating a whole lesson and with the new feature, sometimes within the same sentence or suggested translation. This seems to be something that has thus far only been reported by me, though.
Please improve it by making an option to turn it off… For Japanese the user generated definitions are generally far superior to the AI ones, usually they include the pronunciation. So before I used to be able to press 1 on a new word and get a pretty decent translation with reading… now with the AI translation about the good translation I can no longer do that. Please allow us to turn that off, or put it at the bottom of the suggested list!
This is seriously slowing down my workflow having it in there.
5 Likes
It’s even worse in a sentence mode, where we can now see only AI definitions.
I’d prefer a human definition to be shown first, and use AI only for reference, because it looks like the tech is just not there yet.
Actually the sentence translate already does for me the “context” part, if I am not happy with a dictionary one. I don’t feel as if the new feature adds a lot.
2 Likes
I use android. The play store has “Context-Based Meanings” in the What’s New section of the description. This was last updated on the 17th of July, about the time I noticed it.
2 Likes
Seconded - please give me a way to turn this off. This feature is significantly obstructing my use of the Lingq tool
3 Likes
Unfortunately I missed this post. I too am finding the AI translation suspect. My post is here: AI unreliable in iOS app
3 Likes
I think it’s getting better. I think it’s improving, and I think in particular it is better able to define German verbs the root of which appears early in a phrase and a prefix at the end of the phrase.