Skip to content

Feature or bug? #494

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
mircealungu opened this issue Aug 23, 2024 · 1 comment
Open

Feature or bug? #494

mircealungu opened this issue Aug 23, 2024 · 1 comment

Comments

@mircealungu
Copy link
Member

mircealungu commented Aug 23, 2024

From my exercises:
image

This is incorrect, "got" is the meaning of "har fået" not of simply "fået", cf. https://en.wiktionary.org/wiki/f%C3%A5#Verb.

We could

  • make it very clear to the learners that they have to investigate on their own
  • provide easy links to external dictionary sources
  • go NLP heavy and automatically extend user translations to expressions when this is possible, in our case, we would have detected that "har fået" belongs together and although the user clicked only on "fået" we would have extended the translation. however, this will also fail... and then what do we do?
  • maybe don't forcefully extend, but rather, by pre-processing do a very subtle underline under what we think are expressions (collocations?) such that the user can then extend the translation on their own?

For now, I've fixed it for myself:

image
@tfnribeiro
Copy link
Contributor

This would be caught by a dependency parser - which for the languages we currently have users for we can find both rule based and neural based parsers that would perform reasonably well (I used one in my thesis).

In this case I would say less is more. I think Zeeguu current setup should be focused on learning vocab - and grammar should maybe be more of an after thought, either complemented by curiosity or by teachers in the classroom setting. While we could add more resources, I believe you have stated before that that will just be an overload for a user - if they go to a dictionary and then spend time reading there, they might "lose" track of what they were doing. We also have to ensure every language we add, we find good resources to take the users to.

I think if we want to go more grammar heavy, then we have to consider some exercises where users type full sentences or order words, because then you will be indirectly testing these skills - currently, all our exercises are focus on lexical knowledge - which I still think it's more in line to how Zeeguu is setup at the moment.

go NLP heavy and automatically extend user translations to expressions when this is possible, in our case, we would have detected that "har fået" belongs together and although the user clicked only on "fået" we would have extended the translation. however, this will also fail... and then what do we do?

This in particular, is a bit of too generic of a problem - a dependency parser will always return dependencies between words, so then we would have to define which words we would want to link based on the dependency arc - while some are obvious like AUX Verbs, there might be some that are trickier to know if we would always report or not.

In this particular case, I think if you translate 'har' -> 'have' and you get "got" from 'fået' you can conclude that it's the perfect construction which in English would also be "has got/gotten". So in all, the translation is correct - and the translation should be "has got" - because you can't use 'fået' in the past in Danish (that would be 'fik').

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants