Automated Scoring is a developing technology. The accuracy and reliability of these systems have been proven to be much higher. Besides being a time-and money-saver, there are a number of studies which are being conducted to provide variety in feedback, not only on grammatical issues, but also on semantic related issues. This will reduce not only the paper load of the teachers, but as well as teachers’ assessment related issues. In this paper, we remove those dummy note bins, thereby having 183 dimensions in total. We augmented the input by concatenating it with the first-order difference of the semitone filtered spectrogram. We observed a significant increase of the transcription performance with this addition. Compared to feedforward neural networks, recurrent neural network (RNN) are capable of learning temporal dependency of sequential data, which is the property found in music answer scoring.Also, the Long Short-Term Memory (LSTM) unit has a memory block updated only when an input or forget gate is open, and the gradients can propagate through memory cells without being multiplied each time step. Throughout backward and forward layers together, the networks can access to both history and future of the given time frame. Comparative analyses show that the proposed technique outperforms existing techniques.