Just tried this and dug into the source a bit, the keyword matching system is surprisingly clean. I like how you're using a deterministic symbol map rather than leaning on any NLP or AI. Quick question: are you doing simple substring matches, or is there a preprocessing step for synonyms or phrase grouping? It feels closer to a rule-based expert system than most frontend projects I’ve seen lately. Would be interesting to see the parser modularized for other types of symbolic input.
Great question right now it uses normalized substring matching against a preprocessed keyword index, so phrases like “lost at sea” still resolve symbolically even if they’re embedded in longer sentences. No true synonym handling yet, but I’m experimenting with lightweight alias maps to group similar concepts without fuzzy matching. I’d love to eventually modularize the parser so it can be used for other domains like ritual symbolism or folklore parsing.