Hello,
The task you have described is the most fundamental in machine learning. There are many approaches to it :- logistic regression, neural networks, SVM's.
A true machine learning researcher would not simply do/not do the task. For any machine learning system to be successful, one needs to the data to be operated upon, the trade-off between false positives and true negatives, the algorithm for scalibility and design issues if any.
I have done couple of recommender systems based on natural languages and would gladly do yours but I would like to recommend so many available open source tools like weka,NLP toolkit to build machine learning systems if your data is trivial.
Thanks