This paper presents a raw and noisy fingerprint image recognition system based on natural language processing feature extraction methods and ensemble machine learning methods. The main goal of the proposed model is to reach state-of-the-art classification accuracy, even with the noisy images, eliminate false acceptance rates, and cancel the possibility of recreating a fake fingerprint image from a generated template. To achieve this, we omit preprocessing phase such as application of gradient vectors and multiple filter banks that are typically employed in traditional fingerprint recognition systems. Instead, we employ machine learning methods that classify biometric templates as numeric features. The biometric templates are generated by converting raw fingerprint image into a one-dimensional set of fixed-length codes, which then undergoes stylometric extraction of features further being used for classification. The experimental evaluation shows that the system performs as intended. In addition, the computational and storage costs are significantly decreased with respect to traditional systems, which makes it suitable for use in practical applications.