数据工厂去代工化的战略路径——以数据标注行业为例

Strategic Path of De-OEM in Data Factories: Taking the Data Labeling Industry as an Example

数据标注是数据领域的基础工作,而数据标注行业起步较晚,学界对该行业企业研究匮乏,因此以数据标注行业为研究对象,立足于资源基础观理论与制度理论,在对从中国知网和Web of Science数据库检索到的相关文献回顾、对国外数据标注龙头企业发展实践状况分析的基础上,总结提出数据标注行业转型升级的路径。结果发现:数据标注行业处于快速发展阶段,未来发展趋势将会从注重量转向注重质,追求更高的数据质量、安全性和隐私性;目前中国数据标注行业面临标注效率较低和质量参差不齐、存在数据安全风险、缺乏统一行业规范和标准、人工成本日益上涨等问题,同时经营环境面临行业集中度上升的挑战。基于此,在中国致力发展数字经济的背景下,提出通过实施细化标注任务、提高标注效率、制定行业规范等技术路径,以及扩大企业规模、开发自主品牌、提高研发投入等战略路径,提高中国数据标注行业效率及质量,实现数据工厂的去代加工化,提高研发自主品牌的实力,最终实现由劳动密集型产业向技术密集型产业转变。

Data labeling is a fundamental work in the field of data, but the data labeling industry started late and there is a lack of research on enterprises in this industry. Therefore, based on the theory of resource-based view and institutional theory, through the review of relevant literature retrieved from CNKI and Web of Science database,and the analysis of the development practice of foreign data annotation leading enterprises, this paper summarizes and puts forward the transformation and upgrading path of the data labeling industry. The results show that the data labeling industry is in a rapid development stage, and the future development trend will shift from quantity to quality,pursuing higher data quality, security and privacy;the Chinese data labeling industry is currently facing problems of low efficiency and uneven quality, data security risks, lack of unified industry norms and standards, and increasing labor costs, meanwhile, the operating environment is facing the challenge of rising industry concentration. Based on the research results, in the context of China’s commitment to the development of digital economy, it proposes to improve the efficiency and quality of China’s data labeling industry by implementing technical paths such as refining labeling tasks, improving labeling efficiency, and developing industry standards, as well as strategic paths such as expanding the size of enterprises, developing independent brands, and improving R&D investment, in order to achieve the de-processing of data factories, improve the strength of R&D independent brands, and ultimately achieve the transformation from a labor-intensive industry to a technology-intensive industry.