Software companies develop a large number of software products cater to the needs of customers in different domains. Each product offers a set of features to serve customers in a particular domain. Over the time, the product features (resp. their implementations) should be improved, changed or removed to meet new demands of customers. Identifying source code elements that implements each feature plays a pivot role in such software maintenance tasks. In this article, we present an approach to support effective feature identification and documentation from source code. The novelty of our approach is that we identify each feature implementation based on a semantic-correctness model that can achieve satisfactory results according to well-known evaluation metrics on the subject. We have implemented our approach and conducted evaluation with a large case study. Our evaluation showed that our approach always achieves promising results.