Abstract: The most salient features of Internet of Things (IoT) are its ubiquitous large scale information gathering and intelligent processing to meet everyone's needs. Current solutions in Chinese topic detection and clustering have high time complexity such as O(n2) or O(n3). This paper presents an efficient and patented algorithm for defining topic detection and information clustering over the Internet of Things by combining an improved unigram language model and full text retrieval technique to reduce the time complexity. The experiments and real world applications show that the new method possesses much lesser time complexity.