OrderWords Exercise

[tfnribeiro]

It would be good to deploy the OrderWords exercise for users, as there are NLP pipelines for our most used languages.

There is the issue to consider that this exercise falls a bit outside the category of training just vocab, but extends to testing an overall knowledge of the language structure.

Possible ways of deploying it are:

1. At the end of an article, we have the review words, but we could also have practice sentence construction. This would pick a sentence with no translations (implies the user is familiar with the words) and they get OrderWord exercises for those sentences. A random word could be used as the "fake bookmark".

This first idea has the issue that everything in our database relating to the exercises is associated with bookmarks, so this would be a departure from that model we currently have. It's unclear how we would model this, it might require associating an exercise with a text_id or similar.

2. Allow users to "re-train" learned words by allowing the words that have been learned to be used in OrderWords. The assumption here is that they take sometime to learn the word + the context the word appears in, so the next step would be to learn how to re-write the whole context given the shuffled wordlist.

This can be an effective way to further develop an understanding of how to form sentences in the language. In this case, we do not have to re-model anything, and instead the exercise can be used as is, using the bookmarks learned as the starting point.

[merleschoen]

I've tried the Order Words exercise last week and found it kinda addictive, so I definitely think we should deploy it :)

I like the idea of having it as part of the final review before a word is learned or, like you suggest @tfnribeiro, to have it after a word has been learned.