I played with it a little bit. > My input: "You can't say cows are not alive." > > Successfully matched on Pattern 46, "You can't say <negated true statement>". > Also matched on Patterns 1367, 697, 740, 1082, and 676. Some of these are > patterns for which the entire content is inside brackets, e.g. "<tesla makes > autopilot cars>" and "<delhi is crowded>" I guess those two can match > literally anything. Are they database lint? > > My input: "You can't say that cows are not alive." > > Same matches as the above, but the connection word "that" gets included in > the extracted wildcard <negated true statement>. > > My input: "You can't say that cows are not rocks." > > Same matches as the above. I guess it doesn't know which statements are true > yet. > > My input: "An apple keeps doctors away." > > Successfully matched on Pattern 1421, "<thing1> keeps <thing2> away". > > My input: "Apples keep doctors away." > > Didn't match on Pattern 1421. > > My input: "A cat is an animal." > > Matched on Pattern 1073 and Pattern 661 (plus the others that match > everything).
I'm guessing your future plans include category checks, to see if a potential wildcard is a valid example of <words in brackets> before matching. Right now I'm curious about how all the patterns (templates) currently in the database were generated. Does user input have the option of becoming new patterns? Did you extract them from a dataset? Or have you and your team manually made them all? ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T29d9a364b5df2085-M03577c0154a900e14f8130fc Delivery options: https://agi.topicbox.com/groups/agi/subscription
