Document Wikitionary usage within Kagi Translate Dictionary

Thibaultmol

Just to be clear what I mean with the title:

When you do a dictionary lookup in Kagi Translate it first checks a real dictionary like Wiktionary to see if it's a known word

if it exists, then copy that and have the llm build all it's other stuff around it
if the word doesn't exist, then do the current full-llm dictionary process

This might avoid some of the hallucinations we occasionally see around definitions.

RoxyRoxyRoxy

CC @Giorgio

Giorgio

This is already a thing, thought of it already 🙂
We should probably document it somewhere.

Thibaultmol

Ideally mention it in the UI itself somewhere that Wiktionary is used and how (it's not being copied exactly, it's being used as input to the LLM to generate it's output?)
And if so, is it the entire wikitionary page or specific parts?

Also if Wiktionary was used, we should also mention that in the sources for the new dictionary widget

Could we use the real recordings of pronunciation (with various versions of a language) from wiktionary instead of the ai voices, when available? (I can make this is a seperate kf if you prefer)

Giorgio

Added a mention in the word card, for using the real recordings of a pronunciation please make another kf - it's not immediately straightforward to implement 🙂 Thank you.