Joint Intent Detection and Entity Linking on Spatial Domain Queries

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 4937–4947 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 4937 Joint Intent Detection and Entity Linking on Spatial Domain Queries Lei Zhang1,2∗, Runze Wang3,2, Jingbo Zhou2†, Jingsong Yu1, Zhenhua Ling3, Hui Xiong4 1 Peking University, 2 Business Intelligence Lab, Baidu Research, 2 National Engineering Laboratory of Deep Learning Technology and Application, China, 3 University of Science and Technology of China, 4 Rutgers University 1801210727@pku.edu.cn, wrz94520@mail.ustc.edu.cn, zhoujingbo@baidu.com, yjs@ss.pku.edu.cn, zhling@ustc.edu.cn, hxiong@rutgers.edu Abstract Continuous efforts have been devoted to lan- guage understanding (LU) for conversational queries with the fast and wide-spread popu- larity of voice assistants. In this paper, we first study the LU problem in the spatial do- main, which is a critical problem for provid- ing location-based services by voice assistants, but is without in-depth investigation in existing studies. Spatial domain queries have several unique properties making them be more chal- lenging for language understanding than com- mon conversational queries, including lexical- similar but diverse intents and highly ambigu- ous words. Thus, a special tailored LU frame- work for spatial domain queries is necessary. To the end, a dataset SMQ was extracted and annotated based on the real-life queries from a voice assistant service. We then proposed a new multi-task framework MELIP that jointly learns the intent detection and entity linking tasks on the SMQ with invented hierarchical in- tent detection method and triple-scoring mech- anism for entity linking. A specially designed spatial GCN is also utilized to model spatial context information among entities. We have conducted extensive experimental evaluations with state-of-the-art entity linking and intent detection methods, which demonstrated that MELIP can outperform all baselines with a sig- nificant margin. 1 Introduction The past few years have witnessed the successful deployment of voice assistants on smart speakers (e.g. Amazon Echo) and mobile devices (e.g. Ap- ple Siri and Google Assistant). As a critical step to facilitate informative responses by voice assistants, ∗Lei Zhang and Runze Wang contributed equally to the paper. This work was done when they were interns at the Baidu Research †Jingbo Zhou is corresponding author. language understanding (LU) has attracted tremen- dous research attention in recent years (Wang et al., 2018; Chen et al., 2019; Haihong et al., 2019). LU typically includes the intent detection which de- tects the categorical intent label, and the slot filling which indicates the slot type mentioned by certain words (Zhang et al., 2019; Liu and Lane, 2016). In this paper, we first investigate the LU prob- lem in the spatial domain. With the continuous improvement of their intelligence, virtual assistants are designed to provide many location-based ser- vices such as recommending restaurants (Luo et al., 2020) and providing route planning (Chen et al., 2013). We name all such queries as spatial domain queries that usually contain some spatial informa- tion. Similar to the LU of common conversational queries, there are also two main tasks for the LU of spatial domain queries. The first part is intent detection which aims …

Content truncated to 500 words