Why is "ducks" sometimes "lachan" and sometimes "lachain"? Seems like it's usually "na lachan" and otherwise "lachian".
"Lachan" is a typo in many of these sentences that we are trying to get removed.
One year later, this sentence remains uncorrected. Is there something that’s needed from the Duolingo infrastructure people to be able to resolve the problem?
Yes, you see, when a course is released, the tree is locked. Contributors cannot add or delete skills, lessons or words/lexemes (we can only add, change or delete individual sentences, and tips and notes). This is done because Duolingo bases a lot of their work on metrics obtained from users using the course. If we were allowed to change words, then the course itself would be changed, and this would cause a lot of problems: (1) Existing metrics may become invalid (2) It means users who have already completed some of the course would have to go back along the tree and re-learn some skills, which is awkward when done for small changes (3) Sentences containing the word later in the tree are deleted, meaning some words might not have the minimum of three sentences, and new sentences containing the new word will start appearing, even though the user hasn't learned it yet. It just gets...messy.
That's why we have tree versions (like Tree 2.0 that we are working on at the moment). Tree versions allow us to edit words, skills, lessons, etc. as we could when we were in alpha. This allows us to make a lot of changes at once, while making sure that all words have enough sentences, etc. It also gives us an opportunity to test the tree with a percentage of our users before releasing it to all our users, unlike in phase 1.
Even though changing something like "lachan" to "lachain" seems like a small change, Duolingo doesn't make exceptions like that, and so it's a change that can only be made with a new tree version (and, btw, it has already been fixed in Tree 2.0).
Thanks for the peek behind the curtain. When a tree version is released, are all inflected forms of each lexeme used in that tree version included as part of the release, or are only those inflected forms used in that release’s sentences included?
It makes sense that skills and lessons are frozen with a release, but also freezing lexemes prevents the correction of unintentional typographical errors. Given the variety of courses here, I’d think that the Duolingo infrastructure people would want each inflected form in each language to be uniquely identifiable, so that e.g. the English pronouns “I” and “me”, the Irish prepositions i and in, the Italian articles i and gli, the Polish conjunction i, etc. could all be distinguished from each other (say through the use of behind-the-scenes unique numeric IDs), so that typographical errors could be corrected without affecting the metrics data, which would be gathered via the unique IDs rather than via the inflected word forms. But that’s a matter for the Duolingo infrastructure people, and I appreciate that that’s not something within your control.
When a tree version is released, are all inflected forms of each lexeme used in that tree version included as part of the release, or are only those inflected forms used in that release’s sentences included?
All forms of the word/lexeme that were added to the course are included, even if they are not used in any sentences. For example, in this lesson from the Animals skill, all the words were added individually with no forms, so all of these words appear in sentences. In this lesson in the Past 2 skill which is later in the tree, each word was added as a "forms of". This means that Duolingo counts it as one word and we have to manually add each form of that word. Each form does not have to be included in a sentence but every form is still available when a tree version is released (it doesn't have to be included because it is assumed that students do not have to be taught that form specifically and if they were to see it in a different skill or outside of Duolingo, they could deduct the meaning themselves). The issue with lachan/lachain is that it is included in an early lesson in the tree and Duolingo does not allow us to add "forms of" words so early in the tree (they found it is best to teach forms individually). That means that not all forms of lacha were added to the tree (only two forms were added individually: lacha in Animals, and lachan in Plurals, which has now been corrected to lachain). And since the correct nominative plural form of lacha was not able to added in the convenient location of the Animals skill, we were not able to fix this issue until now. In later skills, all forms of words are added (eclipsed, lenited, genitive, plural, singular, etc.) even though not all forms of all words are actually used in the course.
TL;DR: All inflected forms of lexemes that have been added to the Tree are included in its release, regardless of whether or not they are included in a sentence.
One peculiarity with lexemes on Duolingo is that there are two systems in place: in-house (courses teaching the "in-house languages": Spanish, French, German, Italian, Portuguese and English. These are languages whose original course was developed by Duolingo themselves) and Incubator (courses teaching "new" languages like Irish, Dutch, Danish and Swedish. These courses were originally created by contributors). The in-house courses have a sophisticated lexeme tagger system, where every word added is tagged with information like number, grammatical gender, tense, part of speech, etc. Meanwhile, Incubator languages have a very basic system for adding words.
(to alex, but posting here due to nesting): again, that's for the behind-the-scenes peek. It's too bad the tools aren't more flexible.
At this point I'm actually kind of thankful for this mistake; when I first came across it almost a year ago it completely confused me--I knew it was wrong but I didn't know how to derive the correct forms from it. But having learned more of the language I appreciate this sort of plural/genitive confusion. Duo doesn't really give you many opportunities to correct grammatical errors (arguably the "pick from three" questions, but not really), so it was a nice exercise. :D
So, when does tree 2.0 get released? I just started learning about 2 weeks ago.
Might be an idea to fast forward acceptance decision of an exception once 2 contributors have flagged an exercise.
There are currently only 4 "contributors" for the Irish course.
Reports from users have to be filtered, as the bulk of them are junk - people who haven't a clue file plenty of reports. In cases like this, as alexinireland (one of the original course contributors for this course) explained above, there are structural issues within Duolingo that can prevent certain fixes.
I think "lachan" is genitive, which would mean that it should be "lachain" in this sentence.
I absolutely love these little pearls of surreality. The ducks should not only read the paper but play camogie as well.
It does actually make you sit up and pay attention when there's something like this. I think the vocab and constructions become more memorable this way - and a touch of humour is no bad thing.
From a novice's p.o.v. it looks like the "correct" answer incorrectly uses the genitive "lachan" instead of the correct nominative plural "lachain". Maybe all the detail below is trying to say this.
Yes, and going into detail about why it's surprisingly difficult to fix it now.
Does this sound like ‘iéann’ (which I'm not sure is even possible in Irish) rather than ‘léann’ to anybody else?
Yes, I think that's the pronunciation of the slender 'l' because of the slender vowel next to it.
Is there any place where a user can see a complete list of the words used in a course, incl all inflections?