In recent times, the integration of artificial intelligence into various domains has proven to be transformative, and deep learning models play a pivotal role in solving complex mathematical problems. This research presents a recent contribution to the domain of AI with the development of a specialized dataset tailored for financial word problems. Enhancing an AI-based toolkit for mathematical word problems (MWP), our dataset for financial word problems (FWP) is meticulously curated to train deep learning models focused on solving complex financial scenarios. Motivated by the increasing demand for intelligent systems capable of handling real-world financial problem-solving, our dataset addresses the scarcity of specialized resources in this area. The financial word problems are included in the dataset in various forms, such as finding simple or compound interest, time duration estimation, and principal amount calculation, which are basically applicable in banking sectors. The financial word problems within the dataset are formatted in a corpus structure, enabling the extraction of pertinent information required for solving each problem. The questions posed in the dataset prompt various deep-learning models to determine the results of financial word problems. A comparative study of different models is included in this paper, and we found that the LSTM performs better than other methods in our proposed financial dataset. The dataset can be accessed at https://github.com/samarthita/FWP .