Context dependent AI word translation, new stats page, new themes and more

[PeterBormann]

It’s hard to invest time in learning a new, random platform with all this > uncertainty.

I think good SLA knowledge comes first, AI 2nd for SLA businesses.
That’s IMO definitely not the case with “word-for-word” equations, for example. That was a terrible idea decades ago and it still is…

I think the problem with genAI nowadays is that they’re evolving incredibly fast - and what was great yesterday (say, ChatGPT 3) isn’t great today with ChatGPT 4 omni, etc. That’s probably also a problem for AI-first SLA businesses because you might have to reinvent your business model constantly (i.e. adapt it to the new AI realities)…

[gmeyer]

Personally, I spend more time in each of YouTube (with Language Reactor), Netflix (also with Language Reactor), ChatGPT, and on Zoom with Lingoda than in LingQ.

It’s really only Babbel that’s behind LingQ for me for time spent.

Here’s what helps me with what:

• Reading Comprehension: LingQ
• Listening Comprehension: Netflix and Youtube, with Language Reactor
• Oral Production: Lingoda
• Written Production: Word and ChatGPT

Sliced a different way:

• Low register passive vocabulary: Netflix and Youtube
• High register passive vocabulary: LingQ and some podcasts
• Active vocabulary: ChatGPT and Lingoda
• Grammar: ChatGPT and Lingoda then Babbel
• Pronunciation: Lingoda
• Cultural awareness and formation: Youtube, Lingoda, and Netflix

Once I get to certain milestones in French and in German, I’m dropping Lingoda and moving to Babbel Live. It’s not as good, but a lot cheaper, especially for two languages at once. I might have to add back in iTalki for pronunciation coaching in German.

I remain a fan of LingQ, but it’s very much in the context of their potential.

That said, as someone who’s attained B2 “fluency” in a language nearly 100% online, I can’t imagine doing it efficiently with LingQ as the sole or even primary tool.

In a lot of ways, LingQ’s challenge is that it’s both too big and too small. It’s not “big enough” to do everything. For instance, the online Tutors don’t have critical mass vs such as iTalki and the Writing Exchange can’t compare with what can be done with ChatGPT. Some of the half-baked investments have gotten them no rewards. Yet on the hand, they’re vulnerable to slimmed-down assisted readers that do that one thing sufficiently well and not aspire to be more or need to be more in their business model. I’ve seen other software companies refer to this as a “identity crisis.” There appear to be business decisions that need made.

I commented openly on some of these things a number of months ago, but haven’t for quite a while. I guess I just thought I would here since I personally am quite glad to see LingQ get more AI capabilities subtly grafted into the experience.

(Note: I call out Netflix and Youtube with language reactor being used within the context of those platforms because I prefer the time spent in entertainment-centric experience. I rarely import Netflix and Youtube content into LingQ. While, I’d rather read imported content within Lingoda, in part because the navigation, advertisements, and other distractions are removed, for videos, I’d rather watch them much more immersively in their native platforms with Language Reactor’s suppplemental approach to UX design.)

[DoubleAG]

Contextual translation is what I wanted ever since I first started using LingQ. I’ve never been a fan of having to pick definitions in order to create lingqs. I would prefer to get rid of saved lingqs and just see the unique contextual translation everytime I click on a word. Perhaps this could be an optional setting since I’m sure most users enjoy creating lingqs.

[gmeyer]

Being required to pick a definition of word you don’t know, on the surface, seems a little backwards anyhow.

[davideroccato]

@PeterBormann

I think the problem with genAI nowadays is that they’re evolving incredibly fast - and what was great yesterday (say, ChatGPT 3) isn’t great today with ChatGPT 4 omni, etc. That’s probably also a problem for AI-first SLA businesses because you might have to reinvent your business model constantly (i.e. adapt it to the new AI realities)…

Yes, it is developing fast and there are too many sharks trying to take advantage of it. I’m not against anything like that, but from my current perspective it’s a waste of time to try all these tools without a specific purpose. I’m working on it!

What would be more important is a reliable, already existing software or a company that I know is credible and that will start to build the right tool with AI. At least I know they will be around for a long time and I can invest time in learning the craft.

@gmeyer

In a lot of ways, LingQ’s challenge is that it’s both too big and too small. It’s not “big enough” to do everything.

The problem is that they should have clarity and leadership. The good and simple thing about LingQ is that they have a philosophy, they are input-based learning software.
From my point of view, they should have the courage to drop everything that is not related, because there are hundreds of other tools that can do every other thing right now, and just focus on input-based, but do it well.

They don’t believe in flashcards, they should get rid of them. They should probably get rid of writing exchange, and anything else that is not related to their core philosophy. And improve everything else.

Of course, people are always trying to demand and ask for a lot of different things, that’s why leadership is important. Once you establish your vision and mission, everything else should follow with clarity and purpose.

As you said, we have a lot of tools. I use AI, Language Reactor and many different platforms. From LingQ I look for reliability and excellent reading/listening tools. I don’t even do the listening on LingQ anymore.

If there are too many bugs, go back to basics and do it well, or they will be fagogitated. IMHO

[David72]

Thanks SeoulMate,

It’s good to see we will have the option soon to arrange the dictionary definitions how they used to be.

[bamboozled]

It’s still early days for the ‘contextual hints’ and I fully expect this feature to change (and keep changing) in the future.
The current ‘contextual hints’ look a bit sparse to me, i.e. they are often just one word, even when dealing with an idiomatic expression for example. My experience has been that it’s quite difficult to find one to one correspondence across languages. Sometimes words have a range of possible meanings or simply have no direct equivalent in another language. That’s why I like a good dictionary definition, one that covers a wide range of meanings, possibly even including examples, cultural explanations or usage hints. Additionally I like to save a transliteration in the meaning field, at least in Chinese languages. For illustration, I’ve added a screenshot to show the difference between one of my meanings and the new context hint offered on web.
Of course creating such definitions takes time and consulting online dictionaries disrupts the flow of learning. In recent months I have enjoyed the experimental ‘GPT hints’ feature on iOS, this give useable, nicely formatted results (with the occasional hallucination). In contrast to the contextual hints, these typically contain multiple meanings and in my opinion strike a better balance between a real dictionary and an autogenerated Google translate hint. Since they appear almost instantly, I can just tap words as I go, the app automatically fills it in and saves the GPT hint when I select the next word. Maybe some inspiration could be taken from this feature?

[gmeyer]

David, reacting to your suggested strategy and adding a bit of my own thoughts might be something like:

Mission: Be the leading input-based language acquisition platform.

1. Reading input:
   a. LingQ-created content: Improve the Mini-Stories, to have all the vocabulary and grammar exposure to get through A1 and A2 levels; use generative AI to help improve them
   b. LingQ-curated third-party content: Better curate content for all levels (for instance continuing ahead for such as NHK World for Japanese, RFI for French, DW for German, etc.)
   c. User imported content: Improve the ability for users to better evaluate content (% unknown words especially) rapidly in the import process

2. Audio/video input:
   a. LingQ-created content: None
   b. LingQ-curated third-party content: Ensure LingQ is the easiest place to discover and track progress through content created by language acquisition content creators aligned with comprehensible input (for example Français avec Pierre, French Mornings with Elisa, Français avec Fluidité, Le French Club, Easy French, etc.); Spotify let’s me know when my favorites here have a new episode, LingQ should, heck LingQ should even have tie-ups with these creators where they can “host” the free and paid content on LingQ so they don’t have that complexity/expense and LingQ gets the traffic
   c. User imported content: Improve the ability for users to better evaluate content (% unknown words especially) rapidly in the import process
   d. In-platform experience plug-ins: For Youtube and Netflix, LingQ needs plug-ins like Language Reactor that provide an experience within those platforms, where the audio/video can be consumed directly, with an experience in ways similar to LingoPie too

3, Everything else:
a. Forums here: Sunset it
b. Writing exchange: Sunset it
c. Flash cards and spaced repetition: Either sunset it, or externalize it to an Anki with some sort of API or other integration
d. Writing exchange: Sunset it
e. Tutors: Sunset it
f. Gamification: Either sunset it or greatly simplify the challenges, coins, etc. Despite the disappearance of avatars several years ago, it still remains a relic of the past

Finally, for the non-auto-didactic in the most popular languages, build “courses” out of this content. Uses your own content and third-party content to build courses of lessons that map to levels A1 through C1.

David, I wonder if something like this would be close to what you suggest.

[Tamarind]

I keep looking for the option to deactivate the combined hints and not finding it. Hope it arrives soon!

In the meantime I find that clicking on the one combined overlapping hint will separate the multiple entries to read as they were before the update.

@David72

Thanks SeoulMate,

It’s good to see we will have the option soon to arrange the dictionary definitions how they used to be.

[Tamarind]

My experience has been that it’s quite difficult to find one to one correspondence across languages. Sometimes words have a range of possible meanings or simply have no direct equivalent in another language. That’s why I like a good dictionary definition, one that covers a wide range of meanings, possibly even including examples, cultural explanations or usage hints.

Couldn’t agree more. My study methods have not changed at all. AI and contextual hints may some day in the future work for Korean, but for now I usually put together my own definitions. Examples:

[Obsttorte]

The upside of this approach is that it requires you to think about which possible translation fits best in the situation and, in a broader sense, what the suggested translations have in common to really develop a sense of what a specific word is meant to convey. Personally I often add translations that are not listed in the dictionary, but reflect both the essence of those and the context.
Simplifying or speeding up things isn’t equal to improving. If you have to reflect less about the meaning of the words you encounter, I’d bet this slows down your learning process. Whether this or the speeding up due to the context based ai translation weighs more, we’ll have to see.

[davideroccato]

@gmeyer

a. Forums here: Sunset it

I would keep the forum because it’s a support tool, and an excellent community that few have. It brings value and helps lots of customers to better use the software. Plus, they use an external platform, so they don’t have any specific programming to do.

[baguettenjoyer]

@mark, Can these shortcuts be customizable in the future?

Right now these ones are selected, can other “shared by/providers” be selected?

[mark]

@nfera

What about the refresh button rotates which source is used? As in, first the translation is from ChatGPT, if you don’t like that translation, the refresh button translates with Google Translate, and if you don’t like that translation, the refresh button translates DeepL. So the rotation would be: 1) ChatGPT, 2) Google Translate, 3) DeepL.

That is not being done but is something we are considering. We are first working on getting better full text translations using AI. Once that’s in place, hopefully within a few weeks, we can look at refinements.

[relrod]

This is a serious regression no longer having translations for phrases…

[mark]

@baguettenjoyer

Can these shortcuts be customizable in the future?

We’ll see what we can do there.

@Tamarind

I keep looking for the option to deactivate the combined hints and not finding it.

Sorry, this will be in the next update early next week.

Otherwise, with regard to all the comments, we appreciate the feedback and we will continue to work on and evolve this new feature along with other uses of AI. We would rather put out early versions and get actual usage and feedback than trying to perfect things in a lab.

[mark]

@walkau

It would be nice to have it for the Page View if there are multiple errors in translation that need a refresh.

It refreshes the current sentence and the next 2 each time. No plans for full text refresh but we are working on improving AI translation for full text.

@walkau

would it be possible at some point to get a literal, word-for-word translation

Unlikely but that is what we try to provide when you click on each word. The actual meaning of that instance of the word.

[davideroccato]

@mark

Otherwise, with regard to all the comments, we appreciate the feedback and we will continue to work on and evolve this new feature along with other uses of AI. We would rather put out early versions and get actual usage and feedback than trying to perfect things in a lab.

@mark That means we would be beta testers instead of premium customers. You can’t have it both ways. It’s stressful to study every day without knowing what’s going on, having bugs or changes, or having an unstable product.
We are actively learning a language and not having fun playing video games. We can understand a few problems here and there, but not a constantly changing environment.

You should rethink your strategies, one way or another. Whether by increasing communication, involving your customers in new choices, lowering prices, or other solutions. In this era of increasing AI competition, providing a positive customer experience is paramount to your future success or failure.
IMHO

[Obsttorte]

@davideroccato

That means we would be beta testers instead of premium customers.

This. And there was a time where beta testers got paid. Here it is the other way round.

[SeoulMate]

I am sorry, but this is not acceptable. Paying customers are not beta testers. We understand that new features need to be perfected in some way, but far too often updates are being pushed without proper testing and the website ends up being unusable for days on end.

When you added the new themes for instance, your developers did not properly test the various colour combinations and as a result, we ended up with text that was virtually unreadable. Another example that comes to mind, is the pre-populated glossary in sentence mode that was made available on the web app three months ago. Everyone who uses it knows that it’s a problem when phrase lingqs systematically appear at the end of the glossary and not by order of appearance in the sentence. I pointed out this issue several times, but nothing has been done so far.

Which leads us to another major problem on this website. If you keep pushing out new features when they are not quite ready, you end up with a crazy amount of bugs and issues to correct. As a result, you have to prioritise certain things while others fall are not addressed, resulting in frustration on both parts.

New features shouldn’t be pushed out on the live website until they have undergone thorough testing. If they are still in beta mode , they should be clearly labelled as such and shouldn’t be forced upon users until the stable release is eventually rolled out.

This is why Microsoft and Firefox have got their Canary and Nightly channels respectively. Of course, these are big names, but WordPress.org also offers the opportunity to make a staging website available to clients.