Abstract: In recent decades,users tend to get expected entities directly.Unfortunately,traditional search engine can only return some documents related to the key words instead of the entities user expect.What’s worse,most state-of-art entity ranking methods adopt the approach of weight stack by considering some factors related to the entities,and need many priori knowledge to train the weights.This paper extracted several candidate entities from the snippets returned by search engine and exploited the ideology of “Random Walk” to raise a graph-based algorithm,PERA(Probabilistic Entity Ranking Algorithm),to rank the candidates without many priori knowledge.The results of experiments show that the target entity gets a high ranking score.

Key words: Web,Entity ranking,Search engine,Graph