The project should be able to classify whether any sms is worth reading or not. You can use a Naive Bayes Classifier and Apriori. Use features like text length, who sent it, average word length, wether or not it has attachments, what time it was sent etc. Then you have the user categorize his sms history under two classes: "Worth my time", and "Not Worth my time" to train the classifier.
"Worth my time", and "Not Worth my time"
Naive Bayes Classifier, Apriori