As the automotive industry advances rapidly, user needs are in a constant state of evolution. Driven by advancements in big data, artificial intelligence, and natural language processing, mining user requirements from user-generated content (UGC) on social media has become an effective way to understand these dynamic needs. While existing technologies have progressed in topic identification and sentiment analysis, single-method approaches often face limitations. This study proposes a novel method for user requirement mining based on BERTopic and RoBERTa, combining the strengths of topic modeling and sentiment analysis to provide a more comprehensive analysis of user needs. To validate this approach, UGC data from four major Chinese media platforms were collected. BERTopic was applied for topic extraction and RoBERTa for sentiment analysis, facilitating a linked analysis of user emotions and identified topics. The findings categorize user requirements into four main areas—performance, comfort and experience, price sensitivity, and safety—while also reflecting the increasing relevance of advanced features, such as sensors, powertrain performance, and other technologies. This method enhances user requirement identification by integrating sentiment analysis with topic modeling, offering actionable insights for automotive manufacturers in product optimization and marketing strategies and presenting a scalable approach adaptable across various industries.

Unsupervised domain adaptation (UDA) enables training a model on labeled source data to perform well in a target domain without supervision, which is especially valuable in vision-based semantic segmentation. However, existing UDA methods often struggle with accurate semantic labeling at object boundaries and recognizing minor categories in the target domain. This paper introduces a novel UDA framework—SamDA—that incorporates the Segment Anything Model (SAM), a large-scale foundational vision model, as the mask generator to enhance edge segmentation performance. The framework comprises three core modules: a cross-domain image mixing module, a self-training module with a teacher–student network, and exponential moving average (EMA). It also includes a finetuning module that leverages SAM-generated masks for pseudo-label matching. Evaluations on the GTA5 and Cityscapes datasets demonstrate that SamDA achieves a mean IoU (mIoU) of 75.2, surpassing state-of-the-art methods such as MIC-DAFormer by 1.0 mIoU and outperforming all ResNet-based approaches by at least 15 mIoU. Moreover, SamDA significantly enhances the segmentation of small objects like bicycles, riders, and fences, with, respective, IoU improvements of 4.5, 5.2, and 3.8 compared to baseline models.