HOPE speeches convey uplifting and motivating messages that help enhance mental health and general well-being. Hope speech detection has gained popularity in the field of natural language processing as it gives people the motivation they need to face challenges in life. The momentum behind this technology has been fueled by the demand for encouraging reinforcement online. In this paper, a deep learning approach is proposed in which four different word embedding techniques are used in combination with capsule networks, and a comparative analysis is performed to obtain results. Oversampling is used to address class imbalance problem. The dataset used in this paper is a part of the LT-EDI RANLP 2023 Hope Speech Detection shared task. The approach proposed in this paper achieved a Macro Average F1 score of 0.49 and 0.62 in English and Hindi-English code mix test data, which secured 2nd and 3rd rank respectively in the above mentioned share task.