The Speech Acts Dataset is designed to annotate sentences in an attempt to capture the volition or intent of the speaker/writer. When humans speak or write, the information conveyed by the speaker has an intended purpose. That is, we “do” things with language when we speak and write. This annotation captures the intentions (what is being done) of the speaker or writer.
“We are always doing something with our words when we speak:” (source)
“There are several places where a sentence type other than the imperative is used to give a command” (source)
“taxonomy is simplified in order to retain searchability. For example, when Jesus prays ‘Give” (source)
“commissive speech act is one where the speaker commits themself to some action” (source)
“the correspondence between the content of the utterance and reality.” (source)