Goal-Oriented Script Construction
This addresses the need for script knowledge in task-oriented natural language understanding systems, but it is incremental as it builds on existing methods and datasets.
The authors introduced the Goal-Oriented Script Construction task, where models generate sequences of steps to achieve goals, and demonstrated its practicality and challenge for state-of-the-art Transformer models using a multilingual dataset from wikiHow with 18 languages.
The knowledge of scripts, common chains of events in stereotypical scenarios, is a valuable asset for task-oriented natural language understanding systems. We propose the Goal-Oriented Script Construction task, where a model produces a sequence of steps to accomplish a given goal. We pilot our task on the first multilingual script learning dataset supporting 18 languages collected from wikiHow, a website containing half a million how-to articles. For baselines, we consider both a generation-based approach using a language model and a retrieval-based approach by first retrieving the relevant steps from a large candidate pool and then ordering them. We show that our task is practical, feasible but challenging for state-of-the-art Transformer models, and that our methods can be readily deployed for various other datasets and domains with decent zero-shot performance.