There has been a recent explosion in applications for dialogue interactionranging from direction-giving and tourist information to interactive storysystems. Yet the natural language generation (NLG) component for many of thesesystems remains largely handcrafted. This limitation greatly restricts therange of applications; it also means that it is impossible to take advantage ofrecent work in expressive and statistical language generation that candynamically and automatically produce a large number of variations of givencontent. We propose that a solution to this problem lies in new methods fordeveloping language generation resources. We describe the ES-Translator, acomputational language generator that has previously been applied only tofables, and quantitatively evaluate the domain independence of the EST byapplying it to personal narratives from weblogs. We then take advantage ofrecent work on language generation to create a parameterized sentence plannerfor story generation that provides aggregation operations, variations indiscourse and in point of view. Finally, we present a user evaluation ofdifferent personal narrative retellings.
展开▼