I don´t think we have enough data to train and test an AI for Therion but my knowledge of AI is limited.

As far as I know training data must be counted in the thousands. And them there is the need for test data, another time in the thousands.


Regards,

Rodrigo Severo


On Monday, October 21st, 2024 at 11:42 PM, Bruce Mutton <bruce@tomo.co.nz> wrote:

Just curious, if anyone is using AI to streamline Therion usage or solve Therion problems?

Perhaps in conjunction with versioning platform such as github?

For those of us who are not quite coders…

 

I’ve not thought about it until right now, but a few possibilities come to mind:

  • Troubleshooting peculiar issues that affect only one equate of many in a project.  Find possible reasons why a particular equate should crash Therion.
  • Trawling through a project and highlighting where that project is not internally consistent, and proposing changes.
  • Flattening a project that was created highly nested.
  • Drawing initial scrap outlines given a th2 file with xvi centrelines and PocketTopo or SexyTopo sketches as backgrounds.
  • Weeding out blank pages created with a particular export atlas, by adding a layout statement.
  • Optimising the layout origin position of export atlas to obtain an optimum atlas tessellation.
  • Create new metapost symbols.
  • Suggesting Therion source code changes to further develop atlas support for elevations.
  • Suggesting Therion source code changes to fix, or at least allow control of, line map-connection (problem description in https://github.com/therion/therion/issues/412 )

 

I have 21 Therion data repositories, some with many caves, that could be used to train an AI in my particular usage of Therion.  There are others out there on the web that could be used to teach it a range of possible variations.

 

Bruce Mutton