Look-up and Adapt: A One-shot Semantic Parser

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1129–1139, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1129 Look-up and Adapt: A One-shot Semantic Parser Zhichu Lu⇤ Carnegie Mellon University Pittsburgh, PA zhichul@cs.cmu.edu Forough Arabshahi⇤ Carnegie Mellon University Pittsburgh, PA farabsha@cs.cmu.edu Igor Labutov LAER AI, Inc. New York, NY igor.labutov@laer.ai Tom Mitchell Carnegie Mellon University Pittsburgh, PA tom.mitchell@cs.cmu.edu Abstract Computing devices have recently become ca- pable of interacting with their end users via natural language. However, they can only op- erate within a limited “supported” domain of discourse and fail drastically when faced with an out-of-domain utterance, mainly due to the limitations of their semantic parser. In this pa- per, we propose a semantic parser that gener- alizes to out-of-domain examples by learning a general strategy for parsing an unseen ut- terance through adapting the logical forms of seen utterances, instead of learning to generate a logical form from scratch. Our parser main- tains a memory consisting of a representative subset of the seen utterances paired with their logical forms. Given an unseen utterance, our parser works by looking up a similar utterance from the memory and adapting its logical form until it fits the unseen utterance. Moreover, we present a data generation strategy for con- structing utterance-logical form pairs from dif- ferent domains. Our results show an improve- ment of up to 68.8% on one-shot parsing un- der two different evaluation settings compared to the baselines. 1 Introduction and Background Speech recognition technologies are achieving hu- man parity (Xiong et al., 2016). As a result, end users can now access different functionalities of their phones and computers through spoken in- structions via a natural language processing inter- face referred to as a conversational agent. Current commercial conversational agents such as Siri, Alexa or Google Assistant come with a fixed set of simple functions like setting alarms and mak- ing reminders, but are often not able to cater to the specific phrasing of a user or the specific action a ⇤ These two authors contributed equally user needs. However, it has recently been shown that it is possible to add new functionalities to an agent through natural language instruction (Azaria et al., 2016; Labutov et al., 2018). For example, assume that the user wants to add a functionality for resetting an alarm based on the weather forecast for the next day, as demonstrated by the following utterance: “whenever it snows at night, wake me up 30 minutes early”. The user can instruct this task to the agent by breaking it down into a sequence of actions that the agent already knows. • check the weather app, • see if the forecast is calling for snow, • if yes, then reset the time of the alarm to 30 minutes earlier. This set of instructions result in a logical form or a semantic parse for this specific new utterance. However, this approach can be …

Content truncated to 500 words