Abstract
Different languages might have different word orders. In this paper, we investigate cross-lingual transfer and posit that an order-agnostic model will perform better when transferring to distant foreign languages. To test our hypothesis, we train dependency parsers on an English corpus and evaluate their transfer performance on 30 other languages. Specifically, we compare encoders and decoders based on Recurrent Neural Networks (RNNs) and modified self-attentive architectures. The former relies on sequential information while the latter is more flexible at modeling word order. Rigorous experiments and detailed analysis shows that RNN-based architectures transfer well to languages that are close to English, while self-attentive models have better overall cross-lingual transferability and perform especially well on distant languages. © 2019 Association for Computational Linguistics
| Original language | English |
|---|---|
| Title of host publication | Long and Short Papers |
| Publisher | ACL Anthology |
| Pages | 2440-2452 |
| Number of pages | 13 |
| Volume | 1 |
| ISBN (Print) | 9781950737130 |
| DOIs | |
| Publication status | Published - 1 Jun 2019 |
| Externally published | Yes |
| Event | 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2019 - Minneapolis, United States Duration: 2 Jun 2019 → 7 Jun 2019 |
Publication series
| Name | NAACL HLT 2019 - 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies - Proceedings of the Conference |
|---|---|
| Volume | 1 |
Conference
| Conference | 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2019 |
|---|---|
| Place | United States |
| City | Minneapolis |
| Period | 2/06/19 → 7/06/19 |
Bibliographical note
Publication details (e.g. title, author(s), publication statuses and dates) are captured on an “AS IS” and “AS AVAILABLE” basis at the time of record harvesting from the data source. Suggestions for further amendments or supplementary information can be sent to <a href="mailto:[email protected]">[email protected]</a>.Funding
We thank anonymous reviewers for their helpful feedback. We thank Robert Östling for reaching out when he saw the earlier arxiv version of the paper and providing insightful comments about word order and related citations. We are grateful for the Stanford NLP group's comments and feedback when we present the preliminary results in their seminar. We thank Graham Neubig and the MT/Multilingual Reading Group at CMU-LTI for helpful discussions. We also thank USC Plus Lab and UCLA-NLP group for discussion and comments. This work was supported in part by National Science Foundation Grant IIS-1760523.
Publisher's Copyright Statement
- This full text is made available under CC-BY 4.0. https://creativecommons.org/licenses/by/4.0/