Our client had selected an off-the-shelf nearduper that was underperforming on speed and quality, resulting in review delays.
Build a fast and efficient nearduplication utility, thoroughly test it against the competition, and deploy at less cost to the client.
Hydra neardupe was created one week later. The results were accurate, interpretable and ten times faster. Further enhancements included topic modeling for clustering which we now offer as Extended Metadata upon request.