November 14 News: The National Data Administration announced that on the afternoon of November 10, a pilot construction site meeting on high-quality datasets and data annotation was held in Xiamen. The meeting called for accelerating the development of high-quality datasets and the data annotation industry, emphasizing three key priorities:
First, keep pace with AI advancement and ensure a high-quality data supply. Enterprises' demand for datasets in model development has shifted from "whether data exists" to "whether the right, sufficient, and high-quality data is available." It is essential to stay attuned to AI development trends, adhere to demand-driven and scenario-first principles, clarify data usage requirements, and translate real-world scenario advantages into data advantages, model capabilities, and industrial competitiveness.
Second, consolidate the methods for building high-quality datasets, which can be summarized as "one main line, four links, and one closed loop"—that is, taking scenario-driven construction as the main line, focusing on the four links of demand definition, data governance, professional annotation, and quality evaluation, and forming a "data flywheel" application closed loop.
Third, strengthen the supporting role of data annotation in improving dataset quality.
