If anyone is still interested...
I continue to chip away at this as time and enthusiasm permit. Mostly I've been adding in bits and pieces inspired by the PAW that seem to make sense. I'm not going to implement all the PAW features, just some bits.
Jeremy
I continue to chip away at this as time and enthusiasm permit. Mostly I've been adding in bits and pieces inspired by the PAW that seem to make sense. I'm not going to implement all the PAW features, just some bits.
- Modified the system message logic to be closer to text elements. That is, system messages can now be optimised away if not referenced. This is really to simplify things when I want to add features selectively which require additional system messages. I don't want to include these messages if unreferenced. That was a lot of work.
- Added a new text element type "marker" which allows the matched object name to be inserted into a message. This simulates how the PAW handles messages like "I now have the _."
- Added word type to the vocabulary so that words can be classified as verbs, nouns, adjectives, prepositions and values. This is a mash-up of what the PAW and my original skeleton from a few years ago. This allows disabiguation between, for example, a red ball and a green ball. The way the PAW works is interesting and is far less complicated than what I tried. Mostly the order of words isn't important, so GET RED BALL, GET BALL RED and RED BALL GET are all the same thing really. Anything beyond VERB NOUN is (mostly) handled using conditions in CONDACTS (again, PAW logic) which makes sense as most inputs will be VERB NOUN anyway. In this example, GET BALL takes the first matching object if no adjective specified. This is what the PAW does and again, it seems like a pragmatic choice.
Jeremy
Statistics: Posted by EdwardianDuck — Thu Jul 24, 2025 11:59 am