This paper presents a simple and effective approach to domain adaptation, which works well when there is sufficient target data to improve upon the source data alone. The method involves augmenting the feature space of both source and target data and using this augmented data as input to a standard learning algorithm. This approach is easy to implement, requiring only 10 lines of Perl code, and outperforms state-of-the-art methods on various datasets. It is also easily extended to multi-domain adaptation scenarios. The paper discusses two main types of domain adaptation: fully supervised and semi-supervised. The fully supervised case involves using both source and target data, while the semi-supervised case uses only target data. The paper evaluates several baselines, including SRCONLY, TGTONLY, ALL, WEIGHTED, and PRED, and finds that the proposed method outperforms these baselines. The proposed method is based on feature augmentation, where each feature is expanded into three versions: general, source-specific, and target-specific. This allows the learning algorithm to learn domain-specific patterns while still benefiting from general features. The method is shown to be effective in various tasks, including named entity recognition, part-of-speech tagging, and shallow parsing. The paper also discusses a kernelized version of the method, which is straightforward to derive but not used in the experiments. The method is shown to be effective in several tasks, including ACE-NER, CoNLL-NE, PubMed-POS, CNN-Recap, and Treebank-Chunk. The results show that the proposed method outperforms other approaches, including the PRIOR model, on most tasks. The paper also discusses the results of model introspection, where the learned weights are analyzed to understand how the model handles different domains. The results show that the model learns meaningful patterns that make sense in the context of the task. The paper concludes that the proposed method is simple, effective, and can be applied to a wide range of tasks. It also highlights the importance of developing a formal framework for analyzing domain adaptation models theoretically.This paper presents a simple and effective approach to domain adaptation, which works well when there is sufficient target data to improve upon the source data alone. The method involves augmenting the feature space of both source and target data and using this augmented data as input to a standard learning algorithm. This approach is easy to implement, requiring only 10 lines of Perl code, and outperforms state-of-the-art methods on various datasets. It is also easily extended to multi-domain adaptation scenarios. The paper discusses two main types of domain adaptation: fully supervised and semi-supervised. The fully supervised case involves using both source and target data, while the semi-supervised case uses only target data. The paper evaluates several baselines, including SRCONLY, TGTONLY, ALL, WEIGHTED, and PRED, and finds that the proposed method outperforms these baselines. The proposed method is based on feature augmentation, where each feature is expanded into three versions: general, source-specific, and target-specific. This allows the learning algorithm to learn domain-specific patterns while still benefiting from general features. The method is shown to be effective in various tasks, including named entity recognition, part-of-speech tagging, and shallow parsing. The paper also discusses a kernelized version of the method, which is straightforward to derive but not used in the experiments. The method is shown to be effective in several tasks, including ACE-NER, CoNLL-NE, PubMed-POS, CNN-Recap, and Treebank-Chunk. The results show that the proposed method outperforms other approaches, including the PRIOR model, on most tasks. The paper also discusses the results of model introspection, where the learned weights are analyzed to understand how the model handles different domains. The results show that the model learns meaningful patterns that make sense in the context of the task. The paper concludes that the proposed method is simple, effective, and can be applied to a wide range of tasks. It also highlights the importance of developing a formal framework for analyzing domain adaptation models theoretically.