Lelelo1
commented
3 years ago To clarify, the idea is to loop and use semantic match scoring model on each tag/key that are of interest in the Open Street Map, with the search word. It should give a score, where they highest tag/key then can be identified and used in Open Street map query
Use semantic matching I got another idea that should work out fine and also make it so that I don't need a search engine and having to make externs. The idea is to make a semantic match between the search word and the tags stored in local sqlite which is the download format in the Open Street Map taginfo website.
Create a model in PyTorch and use RoBERTa. Try to create most of the solution in haxe, and compile to C. Later on setup up a good flow, and documentation on how to interact with the compiled C, from C# and Swift. Some dependencies might need to be passed in to the haxe solution.
There is also a potential speed and performance problem when analyzing every tag in the local sqlite database on the device, but it should be able at least partially dealt with multi threading - to set the score of the tags.