Quora Some base lines for Question pairs paraphrase identification: Logistic Regression with n-gram overlap features, lexlatent (VEC and SIM) and lexlatent+MLP.