Enable cross-language search by leveraging summariser

marcel001122

Idea is to possibly push the boundaries of existing universal summariser and Kagi translate that we have in Kagi search to create cross-lingual embeddings to map query-document relationships. e.g. let's say you're an Italian grad student studying Amazonian biodiversity. You search "piante medicinali tribù amazzoniche non documentate" (medicinal plants of undocumented Amazon tribes).

Kagi would:

Find Portuguese-language research from Brazilian universities
Surface Spanish oral history recordings from Peruvian elders
Translate & highlight relevant passages from a French ethnobotany text
Warn you: "83% of results are in other languages - toggle if needed"
Generate a summary comparing findings across cultures

The magic here is preserving native-language nuance while making content more accessible to the user - something no search engine currently does, not even with AI. For use cases can be - A Japanese engineer who could accidentally discovers a breakthrough German mechanical design from 1972 that never got translated! or ancient Indian herb that have potency to cure cancers... idk. wherever the documents exist in different languages.

Luis

As a user would you expect to see results in different languages? Using your example, for a query in italian we could then return pages in german, chinese, thai, etc

This might not provide the most useful or intuitive experience unless the user specifically expects such results; your example references research--wouldn't this functionality be more useful in the context of the Assistant?