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

I am learning Danish and translating to English.

1 Like

There are ways to expose features to a just a whitelisted set of customers that are in a “beta tester” group of users to get their feedback before those features are released to the whole client base.

Some users want to be beta testers. Others don’t.

By not having to release to 100% of the customer base at once, you can de-risk much. Obviously it can be over used and result in supportability headaches so there’s a balance to be found.

Perhaps having some capability to agilely do this will help you increase the velocity of getting new AI features out with managed risk of making the changes.

This also relates to LingQ’s broader software quality issues as this is a technique for managing risks associated with change to a code base.

5 Likes

Echoing this sentiment. The constant breakage and regressions has made it so that I can no longer recommend LingQ to my friends (some of whom have already signed up).

We are paying customers, we expect a stable learning platform. If you want us to beta test for you, then you should make that an opt-in thing through the use of feature flags that users can toggle and incentivize it enough that people want to opt-in. For example, I opt-in to beta test GitHub and that has enabled early access to cool features (and, I might add, has never broken my experience on the site).

The status quo here is simply not acceptable.

5 Likes

From the way I undestood what @mark said was more related to agile vs. waterfall software development management, not necessarily related to premium customers being beta testers.

2 Likes

Thanks for the reply. I do understand that a word-for-word translation might be a technical challenge, given that there are issues with the regular translations. Currently the Spanish is still nonexistent and the Hebrew and Greek translations are still dumpster fires. But translating each word individually in sentence view would be a waste of time. I do prefer uploading my own content and learning it in page view. A simple solution, as I have previously posted, would be to have a option in lesson upload to allow inclusion of a translation. I would be happy to do my own translation separately in Google Translate, ChatGPT (which does literal if you ask it) and be given the chance to upload it so that I know it is correct. This would also help out LingQ given that it would save on compute power and time. Just add a CSV upload for lessons like you have for vocab and we could be in business. Cheers!

1 Like

Love this thread :smiley:

I would expect some dodging from the dev, but they clearly stated they’d rather put their *** in the customers *** at their own expense! How cool these guys are! That’s some attitude!

Some of the customers are trying to escape reality for some fantasy world in which it’s some kind of a mistake or miunderstanding, or even waterfall development management.

The eyes of others are half open, but they continue to vote with their shekels for this.

I even had a chat with a guy who doesn’t want to use a well rounded alternative (with AI, solid functionality base, nice responsiveness of UI etc.) just because it’s made by a Russian immigrant in Germany, because all Russian is bad I guess.

Sad thing is that people who truly believe they don’t have a choice kind of deserve this treatment just by allowing/enabling it.

1 Like

You can do this. Just copy and paste your translation into the Translation popup in the lesson editor. The sentences in the translation will get matched up with the sentences in the text.

2 Likes

I’m very interested in the alternative, can you name drop it or DM it?

1 Like

Apparently I cannot, there’s no DM on the profile page for some reason (maybe because I’ve just created profile a couple hours ago).

Advertising competitors isn’t allowed directly here, so I’ll say it this way:

try to write in Google Play Market “books with translation” and you can find it within top 5 apps with a name almost identical to the LingQ. It has a clean and crisp white UI on the preview and a simple “L” logo.

As you find it, then you can google its website by the name and easily find links for any platforms, including web. That’s how a reader with translation should look like and work imo.

3 Likes

This is not a beta feature. This feature works well. The app works well. At no time did I say anything about using users as beta testers. But, in language learning, everyone has their own ideas about what makes a perfect meaning, tool etc… We have to go with what we think since it’s our product. Then, we learn from what people do and say. We did have a beta user group for this feature. But, we need it accessible to a wider audience to really unearth all perspectives. That way, in the case of AI prompts, we may look at making refinements.
That does not mean that the current iteration isn’t a finished product since it definitely is. I’m using it and it works great. We are adding a setting to disable the combining of definitions. There seems to be a bug with long phrases. I apologize for those shortcomings.
If there are other bugs that you find and you can provide detailed steps to reproduce them, we’ll fix them as quickly as we can.

3 Likes

But it doesn’t. It created a very big regression, as phrasal translation is completely broken, and I use that almost as much as I use word translation.

And no beta tester tried to translate a phrase before this went live to everyone? Is that not a thing people do? Perhaps I am just using LingQ wrong.

5 Likes

Short phrases work well. Long sections of sentences don’t and, unfortunately, were not part of our testing process. We will fix that as soon as possible.

1 Like

But only if the formatting of the original text and the translation fit. If one sentence is broken up into several ones for the translation or vice versa, it won’t match.
On this matter it would be nice for YouTube imports that have subtitles to allow for using a second subtitle set as translation. Many YouTube podcasts aimed at language learners often provide subtitles in several languages. But again, the sentence count might not be the same on all of them, so thinking about a way that allows for a bit more control over the subtitles would be great.

In regards to us beeing beta-testers. No, you did not explicitely state this. However, considering that changes are often released in a state that cause your users quiet some headaches and with often no way to circumvent them plus the fact that there are still a lot of issues either not adressed or pending, you may be able to understand that some of us sometimes feel like we are.

2 Likes

Of course, we understand frustration when new features are launched that require changes to the way the app gets used and may have unintended consequences. Believe it or not, it’s not possible to account for the way everyone uses the app. There are so many different ways that people do things. Change in general is frustrating. We try to respond to feedback.

As for a system for a secondary set of translations, this is unlikely although we are working on improving the interface when matching up translations. Perhaps the update will help you when it is released some time in the next few months.

1 Like

So if I click on a word in LingQ and expect to get a translation for that very word, instead for this plus another one or only another one or a mixture of different languages, or if I get a translation that doesn’t even fit the word family, I am using Lingq in a very specific way that you cannot account for?! Excuse me, but isn’t this “click the word and choose translation” approach the essence of LingQ? Or did I get something wrong.

And when you pushed the themes in the past causing the GUI to become unreadable for many users, this was also due to their specific, unaccountable usage of LingQ? Because who could anticipate that people use LingQ for reading?

Seriously. I can understand that not everything works perfectly from the get go and that some issues will only show up after a new feature has been in use for quiet a while. But you are generalising this as an excuse for complaints that have nothing to do with that. And honestly it becomes tiresome to always read “you are using it wrong” when a complaint is brought up.

Acknowledge the issue, see that you can get it fixed, and consider whether there is something you can improve in regards to how you handle extensions and user communication. That’s all that’s been asked for.

That’s good to hear. Thank you.

6 Likes

Hm, TBH, I’ve no idea what you are talking about here…
I just tested the AI translation feature again with French (“Nota Bene”).
And almost all of the translations (of phrases / shorter sentences / collocations) were both truncated and more or less incorrect. So, those AI translations are just vague hints to get a feel for the meaning, but they should neither be lingq’ed nor learned.

At the moment, those contextual AI translations are more or less useless…

“everyone has their own ideas about what makes a perfect meaning”

There is no such thing as a “perfect” meaning, but there are more or less “correct / incorrect” interpretations. I’d say everyone who has studied languages at a university knows that …

The problem are lay people when they try to defend, e.g., contextual or acontextual word : word equations.
I’ve known way too many of them (esp. as students at school plus their parents).
They usually don’t understand how collocations work (see, for ex., the countless collocation dictionaries for many languages and the extensive research on collocations), they don’t understand sentence mining… and they tend to form sentences that seem syntactically correct on the surface, but are often “wrong” (esp. semantically) because they don’t understand how conventionalized word groups are formed.

Examples in
French:
It’s not enough to learn a “verb” (esp. the infinitive), you have to know the whole construct such as “verb” + à vs de, etc.

German:
It*s not enough to know “choose = aussuchen / auswählen / wählen”, but you have to know, for example:

“choose something” could * be translated as:

  • “Suche (Dir) etwas aus / suchen Sie sich etwas aus”
  • “Wähle (Dir) etwas aus / wählen Sie sich etwas aus”
  • “Wähle (Dir) etwas / wählen Sie etwas”

So the loss of info in a (contextual) AI translation such as "choose something → “wählen” is too great to be useful.

Re users as beta testers
Introducing new features is one thing, but that shouldn’t be an excuse for insufficient testing, sloppy QA and poor UI / UX design decisions (who on Earth, for example, came up with the idea for using a plus sign for removing lesson content??).

I really wish your team and you well with LingQ as a business, but how you handle IT stuff often appears unprofessional (and I work in IT myself). For example, the hundreds and hundreds of forum posts “How can I cancel my subscription / downgrade to a free plan?” alone are a reputation disaster, IMO.

Quite apart from the fact that the forum has more or less degenerated into a long list of bug and problem reports. which is just depressing.

It’s gotten so bad that no one I recommended LingQ wanted to
use it even if I gave them a multi-month subscription as a gift
like I did recently.

But OK, mutatis mutandis, for years these have always been the same discussions that lead nowhere. That’s why I’m turning to other SLA alternatives.

I wish you the best of luck and success…

8 Likes

Just give us the option to switch it off in settings and it will keep everyone happy. Is that a simple thing to do?

4 Likes

Just a thought. Since you have introduced an option to merge definitions, I believe the changes should be reflected in the sentence mode glossary for people who have the option activated. At the moment only the first definition appears regardless of whether the setting is toggled on or off. This is just for the web and iOS as the changes have been implemented in today’s Android update.

1 Like

I am happy that many people like the Context based AI word translations, but it is useless for me and is only preventing me from seeing definitions that I might use. Please add a setting that would allow this to be turned off. You said that many people don’t use settings, so this wouldn’t affect most people, but for people like me it is critical to have this option to turn it off. Thank you.

4 Likes

Yes, I agree that would be nice. We will try and get that done.

3 Likes