Syntactic Approach for Efficient Identification in the Limit from Positive Data of Context-Free and Mildly Context-Sensitive Languages

最近的文献表明，具有反映自然语言现象某一方面的特殊性质的上下文无关语言可以从正向数据中有效地学习。作为前面研究的推广，我们的项目提出了从正数据中学习的高效算法，这些算法甚至可以从更丰富的上下文无关语言类以及轻度上下文敏感语言的类中学习，这些语言处理在自然语言中观察到的一些非上下文无关现象。此外，我们的研究项目提出了一些技巧，在回答学习者有限问题的老师的帮助下，从积极的数据中学习更有表现力的语言类别。

Recently the literature showed that context-free languages with a special property reflecting an aspect of natural language phenomena are efficiently learnable from positive data. Generalizing the preceding research, our project has presented efficient algorithms that learn from positive data even richer classes of context-free languages as well as those of mildly context-sensitive languages, which handle some non-context-free phenomena observed in natural languages. Moreover, our research project has proposed techniques that learn even more expressive classes of languages from positive data with the aid of a teacher who answers limited questions from the learner.