Intent Detection and Slot Filling with Low Resource and Domain
Huu Hiep Nguyen, Ngoc Dung Nguyen, Khac‐Hoai Nam Bui · 2022
This study takes an investigation on intent detection and slot filling, which are two critical tasks for spoken language understanding, of low-resource languages, for instance, the Vietnamese language. Specifically, recent joint models for intent detection and slot filling have achieved state-of-the-art performance, which has proved the strong relationship between the two tasks. In this paper, we focus on the problem of the joint models for two tasks on low resource language, in which the data resource is limited, which leads to the problem of similarity between samples (e.g., in the same domain). Particularly, we present a new method for the aforementioned issues by combining two recent state-of-the-art models in this research field such as JointBERT and JointIDSF for Vietnamese languages. The experiment on PhoATIS, the first public Vietnamese dataset, and our custom similarity for intent detection and slot filling, indicate the promising results of the proposed method.