Zero-Shot Cross-Lingual Machine Reading Comprehension via Inter-sentence Dependency Graph

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F22%3A7MQK4E7L" target="_blank" >RIV/00216208:11320/22:7MQK4E7L - isvavai.cz</a>

Výsledek na webu

<a href="https://ojs.aaai.org/index.php/AAAI/article/view/21407" target="_blank" >https://ojs.aaai.org/index.php/AAAI/article/view/21407</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1609/aaai.v36i10.21407" target="_blank" >10.1609/aaai.v36i10.21407</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Zero-Shot Cross-Lingual Machine Reading Comprehension via Inter-sentence Dependency Graph

Popis výsledku v původním jazyce

We target the task of cross-lingual Machine Reading Comprehension (MRC) in the direct zero-shot setting, by incorporating syntactic features from Universal Dependencies (UD), and the key features we use are the syntactic relations within each sentence. While previous work has demonstrated effective syntax-guided MRC models, we propose to adopt the inter-sentence syntactic relations, in addition to the rudimentary intra-sentence relations, to further utilize the syntactic dependencies in the multi-sentence input of the MRC task. In our approach, we build the Inter-Sentence Dependency Graph (ISDG) connecting dependency trees to form global syntactic relations across sentences. We then propose the ISDG encoder that encodes the global dependency graph, addressing the inter-sentence relations via both one-hop and multi-hop dependency paths explicitly. Experiments on three multilingual MRC datasets (XQuAD, MLQA, TyDiQA-GoldP) show that our encoder that is only trained on English is able to improve the zero-shot performance on all 14 test sets covering 8 languages, with up to 3.8 F1 / 5.2 EM improvement on-average, and 5.2 F1 / 11.2 EM on certain languages. Further analysis shows the improvement can be attributed to the attention on the cross-linguistically consistent syntactic path. Our code is available at https://github.com/lxucs/multilingual-mrc-isdg.

Název v anglickém jazyce

Zero-Shot Cross-Lingual Machine Reading Comprehension via Inter-sentence Dependency Graph

Popis výsledku anglicky

We target the task of cross-lingual Machine Reading Comprehension (MRC) in the direct zero-shot setting, by incorporating syntactic features from Universal Dependencies (UD), and the key features we use are the syntactic relations within each sentence. While previous work has demonstrated effective syntax-guided MRC models, we propose to adopt the inter-sentence syntactic relations, in addition to the rudimentary intra-sentence relations, to further utilize the syntactic dependencies in the multi-sentence input of the MRC task. In our approach, we build the Inter-Sentence Dependency Graph (ISDG) connecting dependency trees to form global syntactic relations across sentences. We then propose the ISDG encoder that encodes the global dependency graph, addressing the inter-sentence relations via both one-hop and multi-hop dependency paths explicitly. Experiments on three multilingual MRC datasets (XQuAD, MLQA, TyDiQA-GoldP) show that our encoder that is only trained on English is able to improve the zero-shot performance on all 14 test sets covering 8 languages, with up to 3.8 F1 / 5.2 EM improvement on-average, and 5.2 F1 / 11.2 EM on certain languages. Further analysis shows the improvement can be attributed to the attention on the cross-linguistically consistent syntactic path. Our code is available at https://github.com/lxucs/multilingual-mrc-isdg.

Druh

J_ost - Ostatní články v recenzovaných periodicích

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

—

Rok uplatnění

2022

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Proceedings of the AAAI Conference on Artificial Intelligence

ISSN

2374-3468

e-ISSN

2571-0966

Svazek periodika

36

Číslo periodika v rámci svazku

10

Stát vydavatele periodika

US - Spojené státy americké

Počet stran výsledku

9

Strana od-do

11538-11546

Kód UT WoS článku

—

EID výsledku v databázi Scopus

—