@SAMUEL FALCÓN PULIDO The service which is close to help you with minimal effort is Azure text Analytics.
But, according to your scenario an NLP model would need training on custom phrases to extract similar phrases from a new document. This scenario of training a custom model with Azure text analysis is not possible currently. You can rely on the API directly to extract Named recognition entities and key phrases which is trained by Azure. This will not provide you the entire text "as-is" in your document but gives key phrases along with sentiment.
Another option is to use the Azure Machine learning service to create experiments with NLP models trained on your data and use the resulting endpoint to extract similar data.