This is a heavyweight solution, that should be tried once we've failed with lots of other heuristics.
Fathom would be the way one would solve this if this is web pages.
We should explore the possibility of serializing the ViewNode as an XML page, and running fathom in an invisible webview or a jscontext.
Linked to #393, this could use the more advanced features of fathom to generate rules/weights for simple rules to get something much more robust than something we can parse by hand.
Related to #215, #371
This is a heavyweight solution, that should be tried once we've failed with lots of other heuristics.
Fathom would be the way one would solve this if this is web pages.
We should explore the possibility of serializing the ViewNode as an XML page, and running fathom in an invisible webview or a jscontext.
Linked to #393, this could use the more advanced features of fathom to generate rules/weights for simple rules to get something much more robust than something we can parse by hand.