The advent of software defined hardware peripherals has enabled the creation of software defined GPS/GNSS receivers. Software defined GNSS receivers have allowed both researchers and receiver designers greater flexibility and creativity when implementing new receiver architectures. This flexibility provides for prototyping and deployment of new signal processing techniques and has enabled designers to leverage machine learning in various areas of the receiver processing chain. One such area that may benefit from this new approach is the decoding of Forward Error Correction (FEC) which is classically performed by the Viterbi algorithm. The goal of this paper is to present a decoder based on a Neural Network architecture and to compare its performance to the traditional Viterbi decoding algorithm employed by most modern receivers. The Bit Error Rate (BER) performance of the Neural Network based decoder is compared with both hard and soft decision Viterbi decoders for randomly generated CNAV messages passed through an Additive White Gaussian Noise (AWGN) channel over various SNRs.