tf-idf is the approach of determine relevant documents by the count of words they contain. While this would emphasis common words like ‘the’, tf-idf takes for each word it’s ratio of the overall appearence in a set of documents – the inverse-doucment-frequence. Here I’ll try to give a simple MapReduce implemention. As a little quirk Avro will be used to model the representation of a document. We are going to need secondary sorting to reach an effective implementation.