Automatic assessment of oral Mandarin proficiency based on speech recognition and evaluation

This paper establishes a speaker-independent pronunciation recognition and assessment system for fairly students of Mandarin as a second language. The recognition part is based on HMM (Hidden Markov Models) and improved in the aspect of prosodic model. Making use of the recognition and detection results and corresponding parametric scorings, the machine scoring is performed to evaluate the quality of pronunciation its correlation with expert score are discussed. The correlation between scores assigned by experts and machine scores was determined to assess their suitability as proficiency indicators. Overall, the results indicate that, even for the narrow range of proficiency levels observed in the test population, the machine scores give a fair indication of oral proficiency. Through integrating recognition and assessment system, we ultimately establish an open, shared and growing-up virtual Mandarin learning community, where all of learners, viewers, professors, without any restrictions on time and places, can express their perspective about how to advance Mandarin learning freely.