This paper explores the application of Noise Contrastive Estimation (NCE) for parameter estimation of directional distributions, with a particular focus on toroidal data. Conventional methods for estimating parameters for such distributions often suffer from computational complexity due to the evaluation of the normalizing constant, not available in a closed form. The proposed approach employs contrastive learning to estimate parameters without requiring a direct evaluation of this constant. Monte Carlo simulations demonstrate that NCE provides unbiased estimates and exhibits rapid convergence to the true value as sample size increases, reducing Root Mean Squared Error (RMSE) significantly. These findings suggest that NCE is a useful alternative for estimating directional distribution parameters.