Digitizing 40 Years of Kenya’s Coastweek Archive to Test and Reduce AI Hallucinations
My name is Linet, I work at the Technical university of Mombasa as an eresource and reference librarian. Our University currently holds 40 years of rich historical coastal publicatications(Coastweek) in bound volumes and digital repository(https://coastweek.tum.ac.ke/) This magazines were handend over to our library by the owner in 2021 therefor we have the copyright to digitize them and make them available to the public for research. while we have made efforts to digitize this publications in a local repository, we are seeking to expand the project and create plug and play data sets that can be used as training model for NLP and historical ai models. we also want to build a search tool that link answers to real coastweek stories so that AI models can give verifiable citations and references when researchers are using AI.
Why It Matters
Currently AI models leans much on the western data bias and more than often it hallucinates when given queries about East Africa History or Africa as a whole. Besides preserving vulnarable history for more generations to come,our proposed search tool will ground AI outputs in real verifiable historical research also we will be giving NLP researchers high quality and authentic kenyan historical text that doesnt exist in standard AI training.
Status
While the foundational infrastructure is in place, it is currently only limited to local hosting. Since we have already digitize most of this publications into our local repository, we are ready to add more into the project by ensuring we complete the digitization of remaining bound volumes and creating datasets for NLP training.
There are no bids on this project.
Comments
No comments yet. Sign in to create one!