Question answering is a crucial task in natural language understanding, as it can be applied to a wide range of natural language processing challenges. Recurrent Neural Networks (RNNs) are commonly used as a baseline model for various sequence prediction tasks, including question answering. While RNNs excel at capturing global information over a long span of time, they may not effectively retain local information. To address this limitation, we propose a model that combines both recurrent and convolutional neural networks, allowing for end-to-end training using backpropagation. Our experiments on the bAbI dataset show that this model can significantly outperform the RNN model in question answering tasks.