In many situations large datasets are available but unfortunatelylabeling is expensive and time consuming. Active Learning is a conceptfor building classifiers by letting the algorithm choose the trainingdata it uses. This achieves greater accuracy than just labeling a randomsubset of the available dataset.The active learning algorithm selects some unlabeled data instanceswhich are then labeled by a human annotator. Given this information aclassifier is trained and new instances for the human annotator to labelare selected. This iterative process tries to label as few instances aspossible while achieving high classification accuracy.In this talk I will give a general overview of the core concepts andtechniques of active learning like algorithms for selecting the queriesand convergence criteria.