From: An analytical study of information extraction from unstructured and multidimensional big data
Rule-based approaches | Learning-based approaches |
---|---|
Interpretable and suitable for rapid development and domain transfer [114] | The performance of machine learning approaches is better in terms of precision and recall but appropriate feature selection is important [115] |
Humans and machines can contribute to the same model. So it is easy to incorporate domain knowledge [114] Heavily rely on domain thesauri [11] | Generating training data is time consuming in learning-based approaches whereas rule-based approaches require pre-defined vocabularies [116] |
Although rule-based systems require domain knowledge and are time consuming, results proved that these are more reliable and useful for automated processing [117] | No experts are required and system can be developed quickly with relatively low cost [118] |
Declarative [119] | Adaptable [119] |
Requires tiresome manual work [118] | Less manual effort [118] |
Highly transparent and expressive | Higher portability than rule-based [9] |