Duplicate question detection in question answer website using convolutional neural network

Online forums are platforms for gathering, sharing information, and discussing between users on a particular topic. Users in online forums can ask questions about a topic, then other users who are experts on that question would answer the question. However, because users can ask questions in various ways, sometimes they ask questions that other users have previously asked. Therefore, a model is needed to detect the semantic similarity of questions in online forums. In this study, we are using Convolutional Neural Networks (CNN) to detect the semantic similarity of questions. To capture the semantic similarity between questions, we are using Glove pre-trained word embeddings. This word embedding vector is used as an input for CNN, then the output is compared with Siamese Neural Networks. Model Optimization is done using Stochastic Gradient Descent. Our model can achieve accuracy of 79 which proved to be higher than Jaccard Similarity and Multilayer Perceptron algorithms. © 2019 IEEE.