Abstract  In this paper, we proposed a novel method to detect  new words in domain-specific fields based on user  behaviors. First, we select the most representative  words from domain-specific lexicon. Then combining with user behaviors, we try to discover the  potential experts in this field who use those terminologies frequently. Finally, we make further efforts  to identify new words from behaviors of those experts. Words used much more frequently in this  community than others are most probably new  words. In brief, our method follows a collaborative  filtering way: first from words to find professional  experts, then from experts to discover new words,  which is different from the traditional new word  detection methods. Our method achieves up to 0.86  in accuracy on a computer science related data set.  Moreover, the proposed method can be easily extended to related words retrieval task. We compare  our method with Google Sets and Bayesian Sets.  Experiments show that our method and Bayesian  Sets gives better results than Google Sets