STIL -- Simultaneous Slot Filling, Translation, Intent Classification, and Language Identification: Initial Results using mBART on MultiATIS++
This addresses the problem of reducing development costs for multilingual systems by enabling some components to be monolingual, though it is incremental as it builds on existing models and datasets.
The paper tackles the STIL task for multilingual NLU, which simultaneously performs slot filling, translation, intent classification, and language identification, and finds that using mBART on MultiATIS++ achieves competitive results, with intent accuracy at 96.07% and slot F1 at 89.87% without translation, and only small degradations of 1.7% and 1.2% respectively when translation is added.
Slot-filling, Translation, Intent classification, and Language identification, or STIL, is a newly-proposed task for multilingual Natural Language Understanding (NLU). By performing simultaneous slot filling and translation into a single output language (English in this case), some portion of downstream system components can be monolingual, reducing development and maintenance cost. Results are given using the multilingual BART model (Liu et al., 2020) fine-tuned on 7 languages using the MultiATIS++ dataset. When no translation is performed, mBART's performance is comparable to the current state of the art system (Cross-Lingual BERT by Xu et al. (2020)) for the languages tested, with better average intent classification accuracy (96.07% versus 95.50%) but worse average slot F1 (89.87% versus 90.81%). When simultaneous translation is performed, average intent classification accuracy degrades by only 1.7% relative and average slot F1 degrades by only 1.2% relative.