A terminal for implementing speech control records a correspondence between a first keyword text and a second keyword text. When a user inputs a speech, the terminal sends, to a first server for semantic parsing and logical parsing, the speech input by the user. Then, when the first server returns a parsing failure, the terminal obtains a parsed text returned by the first server, replaces the second keyword in the parsed text with the first keyword based on the correspondence, and then sends, to the first server, the text obtained after the replacement.