Word recognition stands as a pivotal element within natural language processing (NLP) and machine learning, crucial for diverse applications ranging from automatic speech recognition to optical character recognition and from text analysis to machine translation. This paper explores TensorFlow, a robust open-source machine learning framework, to tackle the complexities associated with word recognition. The research introduces innovative methods employing deep learning and neural networks to enhance the precision and efficiency of word recognition tasks. The evolution of word-to-text recognition technologies has witnessed transformative strides in recent years, impacting various industries and applications. This study scrutinizes the performance of a CNN-based word recognition system, evaluating accuracy and time variations with changes in hyperparameters, including the number of hidden layers, epoch size, activation function, and training-testing datasets.