OP replied in a thread above.
> Thanks a lot. Sure, first we create a list of all the individual sentences in the article. Then, we use the TextRank algorithm to rank each of the text sentences. Finally, we select the top 5 most representative sentences from the article.