Summary and Future Directions
摘要
This chapter summarizes the breakthrough of knowledge-augmented NLP techniques in three lines: natural language understanding, natural language generation, and commonsense reasoning. It discusses a few research challenges such as the heterogeneity of knowledge, global-vs-local knowledge, and scaling of knowledge augmentation, as well as future directions such as knowledge augmentation for structured data tasks (e.g., information extraction), faithfulness, and diverse generation.