This paper proposes a method noise reduction of topic specific web pages. The novel method is based on the frequency statistics and a method of deleting duplicate web pages through hash table and inverted index. According to the frequency statistics, those frequent occurrences of clauses will be considered to be noise and be deleted. Then the clauses and their urls will become the hash table's keys and values to search the duplicate web pages. We use four Map-Reduce tasks to ensure the method of web noise reduction and deleting duplicate web pages. The experiment result shows that the algorithm can remove duplicate web pages efficiently and accurately.