A search intention recognition method and apparatus, and an electronic device and a readable storage medium, which relate to the technical field of computers. The method comprises: acquiring a user log related to a training keyword (S101); determining expansion factors of the training keyword based on the user log and existing system object information, wherein the expansion factors comprise at least one of the following: probability distribution of a topic on the training keyword, probability distribution of the existing system object information on the topic, probability distribution of the training keyword and an existing keyword on a topic dimensionality, and probability distribution of the training keyword and the existing system object information on the topic dimensionality (S102); using the training keyword and the expansion factors to train a machine learning model, so as to obtain an intention recognition model (S103); and using the intention recognition model to recognize a search intention of a user (S104). According to the method, a lightweight original training sample can be used to obtain a relatively precise intention recognition classification model, thereby saving on cost, and improving the accuracy of searches.