Robust Zero-Shot Cross-Domain Slot Filling with Example Values

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5484–5490 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5484 Robust Zero-Shot Cross-Domain Slot Filling with Example Values Darsh J Shah*1, Raghav Gupta*2, Amir A Fayazi2, and Dilek Hakkani-T¨ur3 1MIT CSAIL, Cambridge, MA 2Google Research, Mountain View, CA 3Amazon Alexa AI, Sunnyvale, CA darsh@csail.mit.edu, {raghavgupta,amiraf}@google.com, dilek@ieee.org Abstract Task-oriented dialog systems increasingly rely on deep learning-based slot filling models, usually needing extensive labeled training data for target domains. Often, however, little to no target domain training data may be available, or the training and target domain schemas may be misaligned, as is common for web forms on similar websites. Prior zero-shot slot filling models use slot descriptions to learn concepts, but are not robust to misaligned schemas. We propose utilizing both the slot description and a small number of examples of slot values, which may be easily available, to learn seman- tic representations of slots which are transfer- able across domains and robust to misaligned schemas. Our approach outperforms state-of- the-art models on two multi-domain datasets, especially in the low-data setting. 1 Introduction Goal-oriented dialog systems assist users with tasks such as finding flights, booking restaurants and, more recently, navigating user interfaces, through natural language interactions. Slot fill- ing models, which identify task-specific parame- ters/slots (e.g. flight date, cuisine) from user ut- terances, are key to the underlying spoken lan- guage understanding (SLU) systems. Advances in SLU have enabled virtual assistants such as Siri, Alexa and Google Assistant. There is also signif- icant interest in adding third-party functionality to these assistants. However, supervised slot fillers (Young, 2002; Bellegarda, 2014) require abundant labeled training data, more so with deep learning enhancing accuracy at the cost of being data inten- sive (Mesnil et al., 2015; Kurata et al., 2016). Asterisk (*) denotes equal contribution. Research conducted when all authors were at Google Research. Figure 1: Misaligned schemas for flight booking from kayak.com (top) and southwest.com (bottom): slot name depart in the two schemas refers to depar- ture date and departure city respectively, hence models trained on one schema may falter on the other. Two key challenges with scaling slot fillers to new domains are adaptation and misaligned schemas (here, slot name mismatches). Extent of supervision may vary across domains: there may be ample data for Flights but none for Hotels, re- quiring models to leverage the former to learn se- mantics of reusable slots (e.g. time, destination). In addition, schemas for overlapping domains may be incompatible by way of using different names for the same slot or the same name for different slots. This is common with web form filling: two sites in the same domain may have misaligned schemas, as in Figure 1, precluding approaches that rely on schema alignment. Zero-shot slot filling, typically, either relies on slot names to bootstrap to new slots, which may be insufficient for cases like in Figure 1, or uses hard- to-build domain ontologies/gazetteers. We counter that by …

Content truncated to 500 words