The paper evaluates five measures of semantic distance or relatedness in WordNet, focusing on their performance in a real-world spelling correction system. The measures compared include Jiang and Conrath’s, Hirst and St-Onge’s, Resnik’s, Lin’s, and Leacock and Chodorow’s. The evaluation methods involve theoretical analysis, comparison with human judgments, and application in a specific NLP task. The results show that Jiang and Conrath’s measure performs best overall, while Hirst and St-Onge’s measure over-relates, Resnik’s under-relates, and Lin’s and Leacock and Chodorow’s fall in between. The paper also discusses the limitations of each measure and suggests that hyponymy is only one aspect of semantic relatedness, with other lexical relations being equally important. The use of malapropism detection as a testbed is highlighted as an effective method for comparing these measures, providing insights into their strengths and weaknesses.The paper evaluates five measures of semantic distance or relatedness in WordNet, focusing on their performance in a real-world spelling correction system. The measures compared include Jiang and Conrath’s, Hirst and St-Onge’s, Resnik’s, Lin’s, and Leacock and Chodorow’s. The evaluation methods involve theoretical analysis, comparison with human judgments, and application in a specific NLP task. The results show that Jiang and Conrath’s measure performs best overall, while Hirst and St-Onge’s measure over-relates, Resnik’s under-relates, and Lin’s and Leacock and Chodorow’s fall in between. The paper also discusses the limitations of each measure and suggests that hyponymy is only one aspect of semantic relatedness, with other lexical relations being equally important. The use of malapropism detection as a testbed is highlighted as an effective method for comparing these measures, providing insights into their strengths and weaknesses.