The quality of requirements specification is crucial for software development, which directly affects the success of the project, cost control, and final product quality. However, the lack of publicly available datasets due to the domain specificity of requirements specification and the scarcity of related research has posed a challenge for in-depth research. To address this issue, this study designed a requirements analysis simulation experiment on the Educoder practical teaching platform for collecting requirements specification data. Subsequently, the quality of the data was enhanced by performing preprocessing operations such as segmentation, removing stopwords, and structured partitioning. Next, the data were manually labeled and the dataset was expanded using a variety of data enhancement techniques to enhance the diversity and complexity of the requirements specification text. Finally, a requirements specification quality assessment dataset was successfully constructed by rationally dividing the data, which provides solid data support for further research in software engineering.