Automatic Text Recognition from Historical Production Records
Odoma developed a complete workflow to detect, extract and transcribe more than 1 million handwritten serial numbers that were previously retrieved manually from the physical archives of the company.
Relying on our Groundtruth Manager software, we trained a machine learning algorithm designed for our client's complex documents.
Well over 90% of the serial numbers could be extracted with 99.9% confidence.
The Client's teams can now search for a record directly from an internal search engine and retrieve the number, the scan and the complete archival documentation in seconds.
Used by Longines' marketing department on a daily basis, this application is saving a considerable amount of time, improved customer service and increased the reliability of the information transmitted: the firm's archives are now better able to serve its customer support activities.
Odoma is currently (2022-2023) doing the fourth campagne of extraction for Longines.
Full automation of manual processes
Time saved by a factor of 50 to search within archival records
Better actionable information provided to customer support department