The translation systems are often not able to determine the presence of an idiom in a given paragraph. Due to this many systems tend to return the word-for-word translation of such statements leading to loss in the flavor of the idioms in the paragraph. This paper suggests a novel approach to efficiently determine probability of any statement in a given English paragraph to be an idiom. This approach combines the rule-based generalization of idioms in English language and classification of statements based on the context to determine the idioms in the sentence. The context based classification method can be used further for determination of idioms in regional Indian languages such as Marathi, Konkani and Hindi as the difference in the semantic context of the proverb as compared to the context in a paragraph is also evident in these other languages.