Inspiration
We got the idea from the linguistics track ideas list. We decided to expand upon it by making it applicable to translated texts.
What it does
It analyzes a text to see if it was translated. If it was, it then finds the author's native language. It can also compare an ambiguous work to a known corpus to determine attribution.
How we built it
We used Claude's skills and subagents functions to create a plugin with multiple different functions.
Challenges we ran into
We ran into challenges with determining the size of the corpus. It was unable to compare two documents to each other. It needed at least 5 documents in the corpus to determine attribution. We also needed to correct Sherlock after errors with authorship attributions and identifying language of origin, after which accuracy improved.
Built With
- claude
Log in or sign up for Devpost to join the conversation.