Speech acts constitute one of the fundamental units of human communication, and investigations based on authentic language data are indispensable for uncovering their underlying mechanisms. However, canonical corpus-based approaches based on formal search suffer from notable limitations due to the non-correspondence between pragmatic functions and linguistic forms. With the advancement of natural language processing, digital intelligence techniques have opened up new avenues for addressing the long-standing problem of the form-function mismatch in speech acts. This paper reviews the major developments in speech act research within both corpus linguistics and natural language processing, and highlights the synergetic potential between the two. Moreover, large language model technologies are expected to serve as one of the mainstream empirical methodologies in future pragmatics research.