Reverse Indexing
As a researcher in the Family History Technology Lab at BYU, I helped make the Reverse Indexing application 12 times more efficient in both storage usage and execution speed.
For those that aren’t framiliar, indexing is a method of transcribing handwritten historical records, such as censuses, so they can be searched electronically. Our application, Reverse Indexing, would use machine learning to guess the contents of the fields in the census, and humans would then say whether the guess was correct or incorrect, 12 images at a time.
To handle the large number of annotation errors, we then designed and built a pipeline from the existing application to a new iOS mobile app used to transcribe and correct the annotations. These features now support over ten thousand users as they electronically archive millions of records every year.