Achieving Superior
ML Performance
IMPROVED
ML Performance
ONGOING
ML Expansion
iMerit is CrowdReason’s trusted partner for improving machine learning performance and facilitating best practices in data management.
Challenge
CrowdReason focuses on innovating and maintaining high-value automation solutions that leverage robotic process automation, virtualized labor, machine learning, and blockchain technologies. One of these products, known as TotalPropertyTax, uses machine learning to automate information extraction from sensitive customer documentation. Information that was extracted, however, often required correction as the model’s outputs were inaccurate, resulting in CrowdReason’s highly -skilled employees spending too much time manually entering data into TotalPropertyTax.
To increase employee productivity, CrowdReason needed to improve their data extraction accuracy and entry methods, and began evaluating data service vendors who could relieve their employees of this critical but time-consuming work.
That’s when CrowdReason discovered iMerit.
“Our employees were spending way too much time on manual data entry. We needed a partner who could free them up immediately while solving the problem at its source.”
Solution
To immediately relieve CrowdReason employees, iMerit customized an end-to-end data extraction workflow that simplified the data-extraction process into smaller tasks. Instead of CrowdReason carrying out the workflow, iMerit annotators now entered the data themselves. To continually test and improve the algorithm’s accuracy, each document’s outputs were evaluated by three separate iMerit annotation experts. When there was consensus that the extraction was accurate, they would be used to create a training dataset for the model.
In just three months, CrowdReason’s machine learning model began to automatically aggregate these fields instead of iMerit’s annotation experts, effectively resulting in an automated process. From here, iMerit targeted poorer outputs from the model and put these through the same process involving three separate iMerit annotation experts. In the rare case where a consensus could not be reached around the model’s outputs, the case would be escalated to an iMerit team leader along with data including document information, document image, and client.
“iMerit continues to be an invaluable partner for us. They provided us with accurate data early on, which helped us get up and running with the development of our property tax software.”
Result
With this automated process in place, CrowdReason’s employees now spend 80% less time manually entering data. Data within the TotalPropertyTax application was now automatically aggregated, with client data also being automatically extracted. As both data extraction and aggregation were now automated, CrowdReason also benefited from a clean and seamlessly managed database that contained uniform data management practices that could be used for future machine learning algorithms.
Today, iMerit and CrowdReason continue
