If it's just the similarity between the characters or words of the two sentences you can just use the Levenshtein distance algorithm, but if it's asking about similarity in structure or context then you need some NLP techniques.