Multi-features Enhanced Multi-task Learning for Vietnamese Treebank Conversion

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F26%3AXQXXWTRJ" target="_blank" >RIV/00216208:11320/26:XQXXWTRJ - isvavai.cz</a>

Result on the web

<a href="http://dx.doi.org/10.1007/978-981-97-8367-0_6" target="_blank" >http://dx.doi.org/10.1007/978-981-97-8367-0_6</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-981-97-8367-0_6" target="_blank" >10.1007/978-981-97-8367-0_6</a>

Result language

angličtina

Original language name

Multi-features Enhanced Multi-task Learning for Vietnamese Treebank Conversion

Original language description

Pre-trained language representation-based dependency parsing models have achieved obvious improvements in rich-resource languages. However, these model performances depend on the quality and scale of training data significantly. Compared with Chinese and English, the scale of Vietnamese Dependency treebank is scarcity. Considering human annotation is labor-intensive and time-consuming, we propose a multi-features enhanced multi-task learning framework to convert all heterogeneous Vietnamese Treebanks to a unified one. On the one hand, we exploit Tree BiLSTM and pattern embedding to extract global and local dependency tree features from the source Treebank. On the other hand, we propose to integrate these features into a multi-task learning framework to use the source dependency parsing to assist the conversion processing. Experiments on the benchmark datasets show that our proposed model can effectively convert heterogeneous treebanks, thus further improving the Vietnamese dependency parsing accuracy by about 7.12 points in LAS. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

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

Project

—

Continuities

—

Publication year

2025

Confidentiality

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

Article name in the collection

Lect. Notes Comput. Sci.

ISBN

978-981-97-8366-3

ISSN

—

e-ISSN

—

Number of pages

15

Pages from-to

91-105

Publisher name

Springer Science and Business Media Deutschland GmbH

Place of publication

—

Event location

Taiyuan

Event date

Jan 1, 2026

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—