TextRank

TextRank is a popular algorithm for extractive text summarization. It is based on one of the most well-known algorithms of all time: PageRank. It was the first search-ordering algorithm used by Google to order search results. It works on the principle of ranking pages based on the total number of other pages referring to a given page. Similarly, for TextRank, the text units – typically sentences – are ranked by how similar other sentences are to a given sentence. The TextRank algorithm works as follows:

1. Reads and extracts text from documents.
2. Splits text into sentences.
3. Converts sentences into vectors.
4. Converts each word in a sentence into a vector.
5. Finds the vector for the entire sentence – one approach would be averaging the word vectors.
6. Calculates a similarity matrix among sentences. This is a matrix that measures how similar sentences are to each other.
7. Creates a graph from the similarity matrix.
8. Ranks the sentences...