Stardust.ai æ˜Ÿå°˜æ•°æ ®
2 Case Studies
A Stardust.ai æ˜Ÿå°˜æ•°æ ® Case Study
The customer, Z Company, a social media company, faced significant challenges in developing a sensitive sentence detection algorithm. They struggled with acquiring scarce and costly real-world data, accurately analyzing complex and subjective semantics, and iterating a model that could generalize across diverse domains. To build a harmonious online environment, they partnered with the vendor Stardust.ai.
Stardust.ai implemented a sophisticated solution combining synthetic data generation using GPT models, active learning for efficient data labeling, and knowledge distillation with human-in-the-loop iteration. This approach optimized the dataset and model performance. The resulting algorithm achieved a recall rate of over 97%, far exceeding the industry average of 80-90%, and was successfully applied to the customer's project with high accuracy and a low false positive rate.