<p>In the context of the digital economy, data is regarded as a critical production factor, and its pricing is essential for promoting the circulation of data assets, ensuring transaction fairness, and driving data sharing and economic development. In this study, an intelligent data pricing model is proposed, which integrates traditional numerical features with non-traditional textual information to improve pricing accuracy. Traditional pricing models often fail to fully capture the complete value of data assets, particularly by overlooking alternative data such as functional descriptions or textual metadata associated with the data. To address this issue, this paper develops a deep learning pricing framework that leverages the light gradient boosting machine (LGBM) and bidirectional encoder representations from transformers (BERT) for textual analysis, significantly improving pricing precision. Experimental results demonstrate that the proposed model achieves lower prediction errors across various training-test ratios, reducing pricing errors by 63.5% compared to traditional models. These findings validate the important role of non-traditional data in enhancing the accuracy of data asset pricing. Consequently, this study enhances the effectiveness of data valuation and underscores the value of alternative data sources in data pricing. It further highlights the potential benefits of incorporating additional data sources and deep learning techniques to enhance the performance of pricing models in future research.</p>