Intent Detection and Slot Filling with Capsule Net Architectures for a Romanian Home Assistant

Anda Stoica,Mihaela Dînşoreanu,Camelia Lemnaru,Tibor Kadar,Rodica Potolea

doi:10.3390/s21041230

Anda Stoica, Mihaela Dînşoreanu + Show 3 more

Open Access

https://doi.org/10.3390/s21041230

Copy DOI

Abstract

As virtual home assistants are becoming more popular, there is an emerging need for supporting languages other than English. While more wide-spread or popular languages such as Spanish, French or Hindi are already integrated into existing home assistants like Google Home or Alexa, integration of other less-known languages such as Romanian is still missing. This paper explores the problem of Natural Language Understanding (NLU) applied to a Romanian home assistant. We propose a customized capsule neural network architecture that performs intent detection and slot filling in a joint manner and we evaluate how well it handles utterances containing various levels of complexity. The capsule network model shows a significant improvement in intent detection when compared to models built using the well-known Rasa NLU tool. Through error analysis, we observe clear error patterns that occur systematically. Variability in language when expressing one intent proves to be the biggest challenge encountered by the model.

Highlights

Architectures for a Romanian HomeIntent detection and slot filling are the main tasks to solve when approaching the problem of Natural Language Understanding (NLU) in a conversational system
We propose and evaluate a joint approach for intent detection and slot filling for a Romanian home assistant scenario, based on Capsule Neural Networks [2]
The results show that the proposed capsule network approach reaches a significantly higher intent detection performance compared to that obtained via Rasa NLU

Summary

Introduction

Architectures for a Romanian HomeIntent detection and slot filling are the main tasks to solve when approaching the problem of Natural Language Understanding (NLU) in a conversational system. In the sentence “Salut pune temperatura pe 17 grade in bucătărie” (“Hello set the temperature to 17 degrees in the kitchen”) which indicates a request for setting the thermostat, the other relevant information that can be extracted is that of the value (in degrees) to which the thermostat should be set, as well as the room in which this change should be made. This information is conveyed through the sets of words “17 grade” (“17 degrees”) and “bucătărie”

Methods

Findings

Discussion

Conclusion