Generating Video Descriptions with Attention-Driven LSTM Models in Hindi Language

Abstract

This research addresses the existing gap in video descriptions for regional languages, with a particular emphasis on Hindi. Motivated by a thorough review of available literature, it was observed that languages like Hindi are inadequately represented in this domain. Consequently, we initiated the project titled "Generating Video Descriptions with Attention-Driven LSTM Models in Hindi Language" to enhance accessibility and inclusion of Hindi multimedia content. Leveraging advanced LSTM models and utilizing the VATEX dataset, our objective is to pioneer advancements in regional narrative video production. By venturing into unexplored terrain, we not only contribute to the promotion of Indian language and culture but also establish a precedent for exploring narrative films in other regional languages. This research is strategically designed to foster diversity, integration, and propel broader advancements at the intersection of natural language processing and multitasking. Our findings demonstrate that our approach yields competitive performance when compared to state-of-the-art video captioning baselines such as BLEU and METEOR. This signifies the efficacy of our methodology in enhancing the quality of video descriptions, thereby contributing significantly to the field of regional language video captioning.