tyjch / neo Goto Github PK

If you have a concept node, then you should be able to search Wikipedia based off the title of that node for relevant pages and optionally create new nodes from those.

#f5f1697

This code only produces 1 article in the category, looking at the corresponding Wikipedia page shows that there should be 12. This seems to be an issue with the wikipedia library. This guide shows how to do something similar but with MediaWiki API but it is in Cypher.

from py2neo import Graph
from neo.graph.data import Node
from neo.graph.wikipedia import WikiNode

wikinode = WikiNode("Category:Bayesian networks")
wikinode.get_articles()

#f5f1697

The title property should be used as __primarykey__ and the value of title should be __primaryvalue__ since this is how py2neo handles MERGE and CREATE clauses.

This code below should replicate the issue.

from py2neo import Graph
from neo.graph.data import Node
from neo.graph.wikipedia import WikiNode

node = Node("Some Node")
wikinode = WikiNode("Set theory")
node = Node("Some Node")
wikinode = WikiNode("Set theory")

This produces duplicate nodes as shown below.

https://github.com/Tyjch/Neo/blob/f5f16979ed74b19ab6bc7d52b998976a0add979a/neo/graph/data.py#L5-L36

#f5f1697

When init is called in Node or subclasses of Node, it should attempt to merge with pre-existing nodes in the graph first.

py2neo Documentation

https://github.com/Tyjch/Neo/blob/f5f16979ed74b19ab6bc7d52b998976a0add979a/neo/graph/data.py#L5-L36

Incorporate spreading activation for WikiNodes (or other subclasses of Node that are capable of creating other nodes with raw data sources in their properties). It should incorporate multiple ways to fire (e.g. document similarity is over a given threshold value, a given key phrase exists in a node etc.) that allow it to be customizable by the user. Some other parameters may be initial activation value, firing threshold, decay factor, etc. as mentioned in the above article.

#c878f00

Both of these methods should be added to the parent class WikiNode instead as they both have similar implementations. Maybe even convert them into a single method that specifies what type of page to expand (as there might be more types of WikiNodes added later, like lists, glossaries, or portals).

https://github.com/Tyjch/Neo/blob/c878f00dbe7fdb717a60983f539e09072938cbab/neo/graph/wikipedia.py#L76-L89

https://github.com/Tyjch/Neo/blob/c878f00dbe7fdb717a60983f539e09072938cbab/neo/graph/wikipedia.py#L102-L122