This talk will describe new methods for generating Natural Language in interactive systems – methods which are similar to planning approaches, but which use statistical machine learning to develop adaptive NLG components. Employing statistical models of users, generation contexts, and of Natural Languages themselves, has several potentially beneficial features: the ability to train models on real data, the availability of precise mathematical methods for optimisation, and the capacity to adapt robustly to previously unseen situations. Rather than emulating human behaviour in generation (which can be suboptimal) these methods can even find strategies for NLG which improve upon human performance.
展开▼