AGIF: An Adaptive Graph-Interactive Framework for Joint Multiple Intent Detection and Slot Filling

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 1807–1816 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 1807 AGIF: An Adaptive Graph-Interactive Framework for Joint Multiple Intent Detection and Slot Filling Libo Qin, Xiao Xu, Wanxiang Che∗, Ting Liu Research Center for Social Computing and Information Retrieval Harbin Institute of Technology, China {lbqin,xxu,car,tliu}@ir.hit.edu.cn Abstract In real-world scenarios, users usually have multiple intents in the same utterance. Un- fortunately, most spoken language understand- ing (SLU) models either mainly focused on the single intent scenario, or simply incorpo- rated an overall intent context vector for all tokens, ignoring the fine-grained multiple in- tents information integration for token-level slot prediction. In this paper, we propose an Adaptive Graph-Interactive Framework (AGIF) for joint multiple intent detection and slot filling, where we introduce an intent-slot graph interaction layer to model the strong cor- relation between the slot and intents. Such an interaction layer is applied to each token adaptively, which has the advantage to au- tomatically extract the relevant intents infor- mation, making a fine-grained intent informa- tion integration for the token-level slot pre- diction. Experimental results on three multi- intent datasets show that our framework ob- tains substantial improvement and achieves the state-of-the-art performance. In addition, our framework achieves new state-of-the-art per- formance on two single-intent datasets. 1 Introduction Spoken language understanding (SLU) (Young et al., 2013) is a core component of task-oriented dialog systems. It consists of two typical subtasks, intent detection and slot filling (Tur and De Mori, 2011). Take the utterance “Please play happy birthday” for example, the intent detection can be seen as a classification task to classify the in- tent label (i.e., PlayMusic) while the slot filling can be treated as a sequence labeling task to pre- dict the slot label sequence (i.e., O, O, B-music, I-music). Dominant SLU systems in the liter- ature (Goo et al., 2018; Li et al., 2018; E et al., 2019; Liu et al., 2019; Qin et al., 2019) adopt joint ∗Email corresponding. models to model the relation between the two tasks, which is a direction we follow. Though achieving promising performances, most prior work only focus on the simple single intent scenario. Their models are trained based on the assumption that each utterance only has one single intent. Actually, users usually express multi- ple intents in an utterance and Gangadharaiah and Narayanaswamy (2019) shows that 52% of exam- ples are multi-intent in the amazon internal dataset. Nevertheless, the existing trained single intent SLU models fail to effectively handle the multi-intent settings with the original network structure. Ide- ally, when an SLU system meets an utterance with multiple intents, as shown in Figure 1(a), the model should directly detect its all intents (PlayMusic and GetWeather). Hence, it is important to con- sider multi-intent SLU. Unlike the prior single intent SLU model which can simply leverage the utterance’s single intent to guide slot prediction (Goo et al., 2018; Qin et al., 2019), multi-intent SLU faces to multiple intents and presents a …

Content truncated to 500 words