Graph Based Semantically Extractive Tool for Text Summarization Using Similarity Score

In the last decade or so, there been an extraordinary explosion in the volume of data generated from various sources. Necessary information from the available data allows us to adopt computational tools for exploring. This line of study intends to shorten lengthy documents to ensure readers can grasp the content on a deeper level by reducing them to a shorter version of the original document. This is known as Automatic text summarization, a process for shortening of the text from original content and also it should be meaningful and understandable for the readers. The key component of this effort is the unsupervised learning method recognized as the Text Rank algorithm. It identifies and automatically extracts essential text components such as keywords or words and marks them as summaries. However this algorithm’s limitation relies on a semantic similarity by calculating with cosine distance between the sentences often overlook for semantic relation between words or phrases. We improved the results by adopting graph based Page Rank algorithm, accessing the Google Universal sentence Encoder to generate a meaningful sentence Embedding in order to get better similarity score of the sentences. The Lex Rank summarizer, which integrates the Page Rank algorithm, enhances this procedure. Sentences with closer similarity scores are recommended by it. This aims to mitigate coverage issues and enhance the summarization process, ensuring a comprehensive and accurate summary.