Text rank not only seems to provide better entity extraction than state oft the art techniques but shows good quality in summary extracton with the advantage that it requires no training, acting only on the local document corpus. Intuitively, TextRank works well because it does not only rely on the local context of a text unit (vertex), but rather it takes into account information recursively drawn from the entire text (graph). Through the graphs it builds on texts, TextRank identifies connections between various entities in a text, and implements the concept of recommendation. A text unit recommends other related text units, and the strength of the recommendation is recursively computed based on the importance of the units making the recommendation. For instance, in the keyphrase extraction application, co-occurring words recommend each other as important, and it is the common context that enables the identification of connections between words in text. In the process of identifying important sentences in a text, a sentence recommends another sentence that addresses similar concepts as being useful for the overall understanding of the text. The sentences that are highly recommended by other sentences in the text are likely to be more informative for the given text, and will be therefore given a higher score.

www.cs.unt.edu