Abstract: English homograph has two types，one is polyphonic because of different part of speech，another is polyphonic because of different senses.The disambiguation of the former is easy to be handled after part-of-speech tagging，while the disambiguation of the latter is more difficult.In this paper，a homograph disambiguation algorithm is proposed using WordNet.In this algorithm，the authors extract semantic words from taxonomy of homograph of its senses and context words，and then compare the two semantic sets.The pronunciation with the maximum score is selected.