AI Assisted Quora Question Pair Similarity Evaluation Using Binary Classification

Abstract

The domain for the study is social media where the task of identifying semantic similarities among phrases remains a formidable difficulty in contemporary times, primarily due to the inherent ambiguity of natural languages. The study presents a direct approach for identifying questions that share semantic similarities by utilizing the Word embeddings and Convolutional Neural Networks (CNNs) possess various capacities. Furthermore, the research showcases the efficacy of employing the cosine similarity metric for the purpose of properly comparing feature vectors. Our model demonstrates strong performance on the Quora dataset and corroborates the current evidence that Convolutional Neural Networks (CNNs) are efficacious for tasks involving paraphrase identification. The semantic relatedness of a pair of concepts refers to the evaluation, akin to that of humans, of their connection. Comprehending and utilizing the semantic connection between phrases can enable the utilization of the vast amount of user-generated material on platforms like Quora. Quora serves as a platform for acquiring and disseminating knowledge over a wide gamut. This platform serves as a means to inquire and engage with individuals who provide distinctive perspectives and high-quality responses. It enables individuals to acquire knowledge from one another and gain a deeper comprehension of the universe. Given the substantial monthly visitor count above 100 million, it is not unexpected that a considerable proportion of individuals present inquiries employing comparable language. Searchers spend more to locate the best response to their question when there are other searches with similar goals. This makes writers feel obligated to provide variations on the same question.