Starting point: Building a neural network
Our client is an eCommerce Analytics platform. With tens of thousands of data points, the company was looking for a scalable and fast way to tag its data sets. The company started building a dataset with 2 internal resources dedicated to labeling but wanted to improve the latency while ensuring the quality of the output tags. They realized the super.AI solution can help them achieve both objectives.
Product matching with super.AI
We helped the customer categorize products and identify the brands from product items with our NER data program. We used the original labeled data as training data for the additional projects we ran with them.
- The customer wanted to check whether two product listings were of the same product or not
- We worked with them to define step-by-step instructions to reduce the likelihood of errors
- We automatically built a training to make sure that labelers understood the task and could perform it effectively
We were able to deliver a 99.1% accuracy in the POC. Based on this, the customer was able to increase the amount of data they were processing. On top of the significant quality, we delivered the labeled data to the customer ahead of the time. Thanks to the increase of automation, they were also able to gradually reduce the costs.
Using super.AI, we were able to significantly accelerate our labeling efforts. We were worried that increasing the automation would reduce the overall quality of the dataset, but Super.AI was able to deliver human quality at AI speeds. We’ve gotten almost perfect accuracy