Kacem Abida, Fakhri Karray, and Wafa Abida


ROVER, PMI, system combination, error detection, context


Among the most promising approaches for the problem of reducing the word error rate (WER) in large vocabulary continuous speech recognition (LVCSR) is through an intelligent combination of the output of two or more speech recognizers. Recognizer output voting error reduction, also known as ROVER, is a well-known procedure for systems combination. However, it appears that this technique’s performance has reached a plateau, and researchers have been trying hard to improve on that performance, using various advanced computational intelligence approaches. This paper presents a new approach to boosting the performance of the current ROVER, by relying on the semantics and context to trim erroneous words during the combination process. Recent experiments have proven that it is possible to slightly outperform the ROVER technique, despite the high false positive rate of the error detection technique.

Important Links:

Go Back