eXplainable AI (XAI) does not only lie in the interpretation of the rules generated by AI systems, but also in the evaluation and selection, among many rules automatically generated by large datasets, of those that are more relevant and meaningful for domain experts. With this work, we propose a method for evaluation of similarity between rules, which identifies similar rules, or very different ones, by exploiting techniques developed for Natural Language Processing (NLP). We evaluate the similarity of if-then rules by interpreting them as sentences and generating a similarity matrix acting as an enabler for domain experts to analyse the generated rules and thus discover new knowledge. Rule similarity may be applied to rule analysis and manipulation in different scenarios: the first one deals with rule analysis and interpretation, while the second scenario refers to pruning unnecessary rules within a single ruleset. Rule similarity allows also the automatic comparison and evaluation of rulesets. Two different examples are provided to evaluate the effectiveness of the proposed method for rules analysis for knowledge extraction and rule pruning.