The growth of community-based query answering platforms such as Stack Overflow has led a way to raising duplicate questions, creating redundancy and reduced answer retrieval efficiency and content quality. Manually identifying duplicate question is time consuming and also needs experienced users. Traditional methods based on lexical similarity have mostly failed to cover the semantic gap between syntactically varying but semantically corresponding queries. The new system proposes a same semantic matching system that systematizes duplicate question detection utilizing Word2Vec and neural network architectures. This proposed system preprocesses input data and changes textual content into semantic representations applying Word2Vec. These preprocessed data is then processed via Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) models to capture contextual relationships between question pairs. Performance evaluation is done on metrics like accuracy and recall, with analysis across different models. The results provide an expandable, automated solution for optimizing knowledge management in large-scale technicalities.