Evaluating unsupervised learning for natural language processing tasks

摘要: The development of unsupervised learning methods for natural language processing tasks has become an important and popular area research. primary advantage these is that they do not require annotated data to learn a model. However, this makes them difficult evaluate against manually labeled gold standard. Using part-of-speech tagging as our case study, we discuss the reasons render evaluation paradigm unsuitable methods. Instead, argue rarely used in-context more appropriate informative, it takes into account way are likely be applied. Finally, bearing issue in mind, propose directions future work processing.