Semantic Text Matching (STM) is one of the fundamental tasks in the field of Natural Language Processing (NLP) research. However, existing datasets are centered around the English language and lack sensitive word variants datasets for sensitive content detection, which restricts the development of Chinese STM. In this work, we present SWV, a large-scale Sensitive Word Variants dataset, which contains four types: polysemy, close-words, homophones, and abbreviations. Specially, we use an LLM to generate similar sentences for the optimal text representation. To our knowledge, SWV is the first sensitive word variants dataset in Chinese. The SWV can serve as a Chinese corpus. Also, this semi-structured data is a natural annotation that can constitute many supervised NLP tasks. Based on SWV, we present a sensitive word variant detection algorithm based on Sound Shape Semantic code (3S-Code) for Chinese STM. Experimental results show that 3S-Code can considerably boost the performance of our system and achieve significantly better result than previous state-of-the-art methods on four available datasets, namely SWV, LCQMC, AFQMC, and BQ.