Wednesday, March 15, 2017

Semantic graph processing and the fake news bots

This stuff goes back years, taking a break from something else.  The idea is that all processing, on the arbitrary web with subject,object size; organized in nest block for storage on SAL, but not necessary, any tree based store mechanism worked.  I defined the parse routine which would convolve to finite trees, generating a third.  The generator issued the output tree to the chosen,mostly default, graph device that implemented the add,remove, match, .. and a few other graph constructor methods.

So, for example, a generator graph may convert one tree into an html tree where the html page was treated as a spanning tree.  The idea was to use DLL capability, then the general parser pre-oped the graph until it entered the  installed tree device.

Sort of a vague remembrance of what that was about.  But, the news bots do this.  They send on tree out convolving abouttyhe web, collecting possible subject matter which wis returned as a semantic tree.  That tree was then convolved with a more suble lexical analysis toot, ith a tree interface, and it grabbed paragraphs and suggestvheadlines.  Then the final pass was the HTML generator, and suddenly you had news as inaccurate as anything found on the WA Post.

Evidently Jeff got a team that does this stuff, great. We need more options for our news, jazz it up a bit.

No comments: