Unveiling Multilingual Text Summarization: Harnessing Generative AI for Multi-Modeled Image and Text Synthesis

In today’s digital and information-driven world, the volume of data being generated is enormous and with the explosion of online content, news articles, reports, etc. the capability to extract key information quickly and efficiently is becoming essential. With the technological advancements in the fields of Natural language processing and Generative AI techniques, multilingual summarization is gaining prominence. Furthermore, the development of summarization techniques specifically for Indian languages is immensely important as India being a linguistically diverse country has a myriad of languages spoken across various regions. With most of the online content and global knowledge being available in English, a substantial amount of people finds it difficult to consume the content. Hence, in this paper, we propose VistaSamiksha, a multifaceted, multilingual chatbot that carries out the task of multilingual abstractive summary generation designed to process input both in textual and image format. The proposed system can take input from any three languages, namely, English, and popularly spoken Indian languages Marathi and Hindi, and can provide an abstractive summary output in the chosen language. We have experimented with different pre-trained sequence-to-sequence models to find out the most suitable models for building our system and have presented a detailed overview of the models and our approach in this paper. The ultimate objective is to offer abstractive summaries in any of the user selected languages from the three available options, when the input is presented in any of the three languages. This addresses the challenge of consuming large volumes of data, thereby aiding users in accessing information efficiently.