“Sankshepan”—Summarizing Kannada Text Using BART Transformer

Text summarization in natural language processing (NLP) is done by different algorithms to quick transform the text file in meaning summary maintaining accuracy with theme of the text. In this paper extractive and abstractive summarization method and tool named “Sankshepan” has been proposed. First, it transliterates the Kannada text into English, checking the accuracy and constructing an extractive summary (internally) and then abstractive summary. Finally, the summary is translated back to Kannada assuring the accuracy and comprehension of the summarization by classifiers. The proposed model “Sankshepan” has the control over the length of input and output as summarized text. This proposal has two impacts. The latest transformer tool of Python bidirectional and auto-regressive transformer (BART) is used to train the trained models. Other modules of transformer like pipeline, summarizer, and Tokenizer of Huggingface have been used for better accuracy. The input is translated using Googletrans of Python.