Research Data Supporting "Multi-domain Neural Network Language Generation for Spoken Dialogue Systems"

This is a natural language generation dataset collected from Amazon Mechanical Turk used in this paper "Multi-domain Neural Network Language Generation for Spoken Dialogue Systems" in NAACL-HLT 2016. It contains two domains regarding to consumer electronics: laptop and TV. Each file is in JSON format and contains a list of tuples. The three elements are dialogue act (semantic representation), sentence generated from AMT workers, sentence generated from our handcrafted template generator. There are 13K and 7K distinct DA and sentence pairs in laptop and TV domain, respectively. All products are anonymous.