Goals and Ideas

Current Goals

Create interactive knowledge-maps that:

• let users select a start node and end-node and have the path between the nodes highlighted/enlarged
• allow users to check-off nodes they’ve completed (can we provide/link-to quizes that users can take to see if they understand a node: play this arpeggio, what’s the d-separation between node X and node Y, translate “Son las tres y cuarto en punto”, etc)
  • Roger: the nodes should link to whatever publicly available resources we can find on the topics, which includes exercises if they’re available. I think it would be a lot of work to make our own quizzes, though it could be feasible for certain high-priority nodes.
    • Colorado: I agree that making our own quizzes would be very time-consuming—-we should probably not worry about this until later; but it might not be too difficult to link to a third party quiz
• provide a graphical interface that allows users to submit new k-maps
  • Roger: This would be really useful. We’ll need to decide how to handle the user-submitted content. Do we want it to be immediately available Wikipedia-style, maybe in a separate section that’s marked provisional (to set it apart from the curated parts)? Or just have it be a form of suggestions?
    • Colorado: we could make it immediately available with some spam-filters in place, then allow the rating system to separate the good from the bad. I like the “provisional” idea. Perhaps provisional kmaps could be marked with e.g. an asterisk in the search results.

Ideas

• Integrate with Anki or a similar knowledge reviewing program
• give users the option to create an account — save progress, display contributions, etc
• how can we initialize knowledge maps for experts? perhaps it’s possible to use topic models and causality (causal topic models?);
  • Roger: I’m not sure what you mean by causal topic models. I think we need to provide a way to save what people have read (as you say above) as well as a way for people to mark that they’ve taken (say) a course on linear algebra.
    • _Colorado: Sorry, that was poorly worded. I’m referring to the problem of automatically creating (initializing) kmaps via a learning algorithm. Topic models seem a natural starting point for this problem but they typically don’t express any kind of directed relationship between the terms in the topics, hence we would need a “causal topic model”. I’ll think about this problem a bit more — could make an interesting paper (automatic kmap creation). Let me know if you have any ideas :) _
• let users rate kmaps
  • Roger: I think user ratings would be really helpful in general. I’d be more likely to do an exercise if it says “76 out of 80 people found this helpful.”
• let user comment-on provide feedback to kmaps (might be helpful in the same way that recipe sites have some users that post altered recipes: use pistachios instead of walnuts)