Affiliation:
1. Department of Computer Science University of Copenhagen, Denmark. rc@di.ku.dk
2. Department of Computer Science University of Copenhagen, Denmark. rahul@di.ku.dk
3. Department of Computer Science University of Copenhagen, Denmark. hcl@di.ku.dk
4. Department of Computer Science University of Copenhagen, Denmark. dh@di.ku.dk
Abstract
AbstractSemantic parsing (SP) allows humans to leverage vast knowledge resources through natural interaction. However, parsers are mostly designed for and evaluated on English resources, such as CFQ (Keysers et al., 2020), the current standard benchmark based on English data generated from grammar rules and oriented towards Freebase, an outdated knowledge base. We propose a method for creating a multilingual, parallel dataset of question-query pairs, grounded in Wikidata. We introduce such a dataset, which we call Multilingual Compositional Wikidata Questions (MCWQ), and use it to analyze the compositional generalization of semantic parsers in Hebrew, Kannada, Chinese, and English. While within- language generalization is comparable across languages, experiments on zero-shot cross- lingual transfer demonstrate that cross-lingual compositional generalization fails, even with state-of-the-art pretrained multilingual encoders. Furthermore, our methodology, dataset, and results will facilitate future research on SP in more realistic and diverse settings than has been possible with existing resources.
Subject
Artificial Intelligence,Computer Science Applications,Linguistics and Language,Human-Computer Interaction,Communication
Reference79 articles.
1. Cross-lingual sentiment analysis without (good) translation;Abdalla,2017
2. Rewarding coreference resolvers for being consistent with world knowledge;Aralikatte,2019
3. Translation artifacts in cross- lingual transfer learning;Artetxe,2020
4. On the cross-lingual transferability of monolingual representations;Artetxe,2020
5. Neural machine translation by jointly learning to align and translate;Bahdanau,2015
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献