Summarization of Hindi News Editorial with Sentiment Determination

Text summarization is a natural language processing application that is being researched extensively and applied further to reduce the processing time for various long-winded text-based activities. However, as NLP is still in its budding phase, the work is relatively limited to the English language, leaving regional languages rather untouched despite having an incredible following of speakers. Such is also the case with the Hindi language. In this paper, we propose to come up with an effective method of summarisation news articles in the Hindi language. Like the English variant of this application, we wish to emphasize on the important sections of a Hindi news report and summarize it within 60 to 80 words. The summarization technique will try to identify the theme of the news, named entities and numbers, title terms, etc., for constructing a keyword table. This will be further compared against a knowledge base with weighted keywords for ranking the important sentences in the relevant order and finally picking out the sentence most needed for the summary. Our goal for summarizing the Hindi news articles specifically roots from the dilemma that despite these articles being a rich source of opinionated information about various topics, they are often ignored by the readers because of their long-winded nature that makes the useful information lost in the sea of words decorated by winded introductions and linguistic ornaments like idioms. Hence, this system should enable in an effective means of summary for finding useful information along with pruning all such irrelevant details.