commonsense team mailing list archive

[Ken Arnold <kenneth.arnold@xxxxxxxxx>]

Review: Approve
Great! Thanks for the merge. A few nits to pick ;)

1. `from labeled_tensor` winds up importing the module twice. To avoid that, always use absolute paths. Future Python's will make that mandatory.

2. Could you explain what's "very hacky" about your important relations algorithm and any thoughts you have about doing better?

3. In cnet_graph_from_tensor, try `for (concept1, (typ, rel, concept2)), score in cnet.iteritems():`. And of course just don't add unwanted edges in the first place.

4. semantic_network: why the frozensets? And I'm not clear on the distinction between `value` and `type`.

5. get_reachable_vertices should avoid recursing if next_v in found_vertices. And it might be more efficient to use a queue.

-- 
https://code.launchpad.net/~r-launchpad-jayantkrish-com/divisi/analogies/+merge/6868
Your team Commonsense Computing is subscribed to branch lp:divisi.