This chapter focuses on the pattern detection and extraction step in text data commonly called text data mining. I examine some of the literature on natural language processing and propose a method of recovering value from the text of virtual group discussions based on methods derived from the communication field. Then, I apply the method in a case using data from 216 different groups from a virtual group experiment. The results from the case show that higher performing groups are characterized by higher frequencies of acts of dominance and higher frequencies of terms concerning cognition, communication and praise. Higher performing groups were also characterized by lower frequencies of acts of equivalence and lower frequencies of leveling terms and numerical terms. Ways to use this knowledge to improve the groups’ performance are discussed.