SPEECH RECOGNITION BASED ON SYMBOLIC REPRESENTATION OF A TARGET SENTENCE

摘要

Systems and methods for processing a user speech input to determine whether the user has correctly read a target sentence string are provided. One disclosed method may include receiving a sentence array including component words of the target sentence string and processing the sentence array to generate a symbolic representation of the target sentence string. The symbolic representation may include a subset of words selected from the component words of the target sentence string, having fewer words than the sentence array. The method may include processing user speech input to recognize in the user speech input each of the words in the subset of words in the symbolic representation of the target sentence string. The method may further include, upon recognizing the subset of words, making a determination that the user has correctly read the target sentence string.