This project addresses the problem of predicting which of the answers given by users to a question on Yahoo Answers is the best one.
It uses Natural Language Processing and Machine Learning algorithms to analyze the textual content of the questions and answers, but also the expertise of the users in a specific domain and their behavior in the community.
The system achieved a ~21%-27% improvement over the state of the art.
This work was developed during my internship at Yahoo! Labs Barcelona and coded in Java. It was accepted and published in ACM Transactions on Information Systems.