Hello, I’m a new user and I’m still experimenting with the platform,
I love how well it runs and the general concept, but I feel like I don’t understand the implementation at all…
There are a few things I don’t really know how to use, but the biggest issue I have right now is with the AI feature, please help me understand
First of all, using a premium subscription, the contextual explanations given by the AI feel very inadequate, they’re basically giving a dictionary definition, and then explaining the sentence (which most of the time is pretty redundant once you’re given a definition). But it’s not really leveraging any of the advantage the AI has compared to a dictionary, for example it’s not saying whether the word itself is common or not, whether or not it’s used in a conventional way (as part of an idiom for example), and will not point out if the selected vocabulary is misused in that context or if there’s a mistake in the sentence.
Furthermore, if you select multiple words it tries to explain whatever you’ve selected regardless of the validity of that selection, even if those words are from completely unrelated clauses. Not only does this give results that are completely unusable, it also means that the AI does nothing to steer you back in the correct direction. It’s bad enough with Italian, but with something like Japanese where Readlang takes every character to be an individual word it makes the entire feature extremely tedious to use, especially when you know in advance you’re going to get very little out of it.
Did I miss something, can I prompt the AI through Readlang to change its behaviour?
Does a premium plus subscription fix those issues? If it does then I’d be happy to upgrade, so I can experiment with it and compare both at least, maybe to use Readlang as my main reading tool, but as of right now the results are significantly worse than even the free version of chatGPT…
Or does this stem from me misusing the feature altogether?
Feel free to let me know how you personally use it, I haven’t found much footage of people really using it in depth, mostly surface level demonstrations.
Thanks in advance for any remarks or suggestions !
One thing to know, is if you go to your main Settings, there is an option at the bottom called “Ask AI” that gives you a space to ask whatever question you want of the AI, at the bottom of the Explain tab.
I think you’re right that on its own Explain is not going to provide much context-dependent insight. And I think you’re right that it often feels redundant to get “what does it mean in general” followed by “what does it mean here.”
I don’t know how much Japanese you know. But I would definitely only recommend Readlang to a student of Japanese (or Chinese) who knows enough to have an understanding of where the word boundaries probably are. Otherwise you have to wade through too much nonsense as the AI won’t typically tell you that what you highlighted isn’t really a word.
In my opinion, upgrading to Premium Plus isn’t going to do much for your concerns. It’s not more customizable, it’s just a little smarter.
Thanks for the answer, I feel like this adresses my initial question very adequately!
I’m currently mostly reading through my web browser, with a dictionary extension and chatgpt on the side for when I get stuck, it works well but has the drawback of using different tools with their own controls. I was hoping to find a platform that would do all those things with a single interface, but maybe its philosophy and mine aren’t compatible, I’m really having a lot of issues trying to use it. I still love the design though!
Regarding your point on word boundaries I totally agree, but the example I gave of selecting words from different clauses was an extreme one, meant to stress the fact that the AI as it’s configured here does nothing to correct the user, and in fact gives the impression that whatever was taken to be a word or idiom is valid even when it’s clear it wasn’t.
The real issue is that if a word is part of an unknown idiom or set phrase, then a learner is likely not going to know any better, and isn’t by him or herself going to accurately select what needs to be explained.
As an example, I just loaded an english text, and selecting “whether it” instead of “whether it is” leads to the AI explaining the term as if “whether it” by itself constitutes a valid idiom. This is a simple one, but more complex and less common idioms are actually very easy to misread even as an intermediate learner, especially with a target language that is heavily idiomatic and doesn’t share a common root with your own.
Additionally, it seems like all the languages available to use with the AI are just translated from English.
So if I set it to French, it feels heavily wooden because it follows the English original very closely and doesn’t actually produces French sounding sentences.
But most importantly, if I set it to a language I’m learning, it will retain the exact same information as it would for English. So if the English AI output said : “This word means Cat in english”, and you set it to Japanese, it will just translate this and even include the English word “cat” in its definition, instead of actually giving a Japanese definition which would be much more helpful to a learner.
Whether or not the context explanation feels redundant depends a lot on the language for me.
Currently I’m reading an English novel and while I look up some more advanced words or interesting phrases my general understanding of the text is quite high, so the context explanations do not provide that much extra value.
But when I was reading a fairly advanced text in Spanish, I often had some aha moments when reading the context explanation and it helped me better understand how the word works in this specific sentence.
On occasion I noticed that the AI correctly pointed out a mistake in my source text, so that’s definitely possible (I’m on the Premium Plus plan, so it could be related to the used model).
I like to use the Ask AI feature too. I mostly ask about etymology, have it compare the word to a similar one or whether or not the word is commonly used.
The last point is something that I would love to see included in the general explanation (could just be some kind of tag like “common”, “archaic”, “slang”, “literary” or something like that).
Regarding the quality of the explanation: I use it in my native language German and I never had the impression that it was translated from English. Of course it sounds like a typical AI text, but that’s as expected
The Ask AI feature sometimes answers in the target language even when I ask something in German, but the explanations in itself used a language consistently.
