Recursive Template-based Frame Generation for Task Oriented Dialog

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2059–2064 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 2059 Recursive Template-based Frame Generation for Task Oriented Dialog Rashmi Gangadharaiah AWS AI, Amazon rgangad@amazon.com Balakrishnan Narayanaswamy AWS AI, Amazon muralibn@amazon.com Abstract The Natural Language Understanding (NLU) component in task oriented dialog systems pro- cesses a user’s request and converts it into structured information that can be consumed by downstream components such as the Dialog State Tracker (DST). This information is typi- cally represented as a semantic frame that cap- tures the intent and slot-labels provided by the user. We first show that such a shallow repre- sentation is insufficient for complex dialog sce- narios, because it does not capture the recur- sive nature inherent in many domains. We pro- pose a recursive, hierarchical frame-based rep- resentation and show how to learn it from data. We formulate the frame generation task as a template-based tree decoding task, where the decoder recursively generates a template and then fills slot values into the template. We ex- tend local tree-based loss functions with terms that provide global supervision and show how to optimize them end-to-end. We achieve a small improvement on the widely used ATIS dataset and a much larger improvement on a more complex dataset we describe here. 1 Introduction The output of an NLU component is called a se- mantic or dialog frame (Hakkani-T¨ur et al., 2016). The frame consists of intents which capture infor- mation about the goal of the user and slot-labels which capture constraints that need to be satisfied in order to fulfill the users’ request. For example, in Figure 1, the intent is to book a flight (atis flight) and the slot labels are the from location, to location and the date. The intent detection task can be mod- eled as a classification problem and slot labeling as a sequential labeling problem. The ATIS (Airline Travel Information System) dataset (Hakkani-T¨ur et al., 2010) is widely used for evaluating the NLU component. We focus on complex aspects of dialog that occur in real-world Intent: atis_flight Slot-labels: from pittsburgh i’d like to travel to atlanta on september fourth O fromloc.city_name O O O O O toloc.city_name O depart_date.month depart_date.day Figure 1: Flat structures used to represent Intents and slot labels in ATIS. ‘O’ for Other or irrelevant tokens. scenarios but are not captured in ATIS or other al- ternatives such as, DSTC (Henderson et al., 2014) or SNIPS 1. As an example, consider a reason- able user utterance, “can i get two medium veggie pizza and one small lemonade” (Figure 2A). The intent is OrderItems. There are two items men- tioned, each with three properties. The properties are the name of the item (veggie pizza, lemonade), the quantity of the item (two, one) and size of the item (medium, small). These properties need to be grouped together accurately to successfully fulfill the customer’s request - the customer would not be happy with one small veggie …

Content truncated to 500 words