hints feature #581
hints feature #581
Conversation
|
|
||
| sprint_api = Blueprint('sprints', __name__, url_prefix='/api/sprints') | ||
| # this script doesn't work | ||
| # !./get_embeddings.sh |
There was a problem hiding this comment.
Yeah, so that is a jupyter notebook specific thing. In a really setting, we would just run this ourselves on the command line somewhere. Maybe modify this comment to say something about needing to download the embeddings.
apis/sprints_api.py
Outdated
| # greedy = GreedySearch(embeddings_provider, graph_provider) | ||
| # path = greedy.search(start, end) | ||
|
|
||
| beam = BeamSearch(embeddings_provider, graph_provider) |
There was a problem hiding this comment.
So here, what you should do instead of doing the whole search is do a single step, i.e. look at the pages linked from your current article, and then compare them with the embedding of the goal.
Doing a whole beam search is a lot of unnecessary work.
Note this would require modifying the wsbot library, which I can help with if you're intested. The best way I see of doing that is adding a separate get_link_rankings or something to the wsbot lib, that both the Search and external users (such as this API) can call.
There was a problem hiding this comment.
Ah that makes sense I'll look into this. Thanks for reviewing!
There was a problem hiding this comment.
@dqian3 Sorry for the wait, instead of making a function for hints that returns rankings I just made one that iterates through the current page's links and returns the closest word based on a greedy approach. I added code for the greedy search to use this but left it commented out for now. Could you take a look when you get a chance?
| @@ -57,6 +57,12 @@ | |||
| Current Article<br><strong>[[currentArticle]]</strong> | |||
| </div> | |||
| </div> | |||
| <div style="float: right;" > | |||
|
hints feature using wiki2vec pathfinding bot
All bot code taken from https://github.com/wikispeedruns/wikispeedruns-bot