Automatic dependency parsing of Estonian: what linguistic features to include?
In this paper, we investigate the role of linguistically motivated input features and pre-processing in advancing dependency parsing of Estonian. In particular, we focus on parsers that take morphological features as explicit inputs, and investigate the effect of a) training data size, b) the choice...
| Publicado en: | Language Resources & Evaluation Vol. 59; no. 2; pp. 1467 - 1495 |
|---|---|
| Autores principales: | , , , |
| Formato: | Artículo |
| Publicado: |
Springer Nature
Jun2025
|
| Materias: | |
| Acceso en línea: | Ver este registro en EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=hlh&AN=185240066&site=ehost-live header: @attributes: shortDbName: hlh uiTerm: 185240066 longDbName: Humanities International Complete uiTag: AN controlInfo: bkinfo: jinfo: jid: 1574020X 179V jtl: Language Resources & Evaluation issn: 1574020X maglogo: N pubinfo: dt: Jun2025 vid: 59 iid: 2 pid: 237 pub: Springer Nature artinfo: ui: 185240066 10.1007/s10579-024-09779-z ppf: 1467 ppct: 28 formats: fmt: – @attributes: type: T – @attributes: type: P size: 1.4MB tig: atl: Automatic dependency parsing of Estonian: what linguistic features to include? aug: au: Laur, Sven Orasmaa, Siim Eiche, Sandra Särg, Dage affil: https://ror.org/03z77qz90 Institute of Computer Science, University of Tartu, Narva mnt 18, 51009, Tartu, Estonia https://ror.org/03z77qz90 Institute of Estonian and General Linguistics, University of Tartu, Jakobi 2, 51005, Tartu, Estonia su: Machine learning Annotations sug: subj: Machine learning Annotations keyword: Clause annotations Dependency parsing Estonian Morphological annotations ab: In this paper, we investigate the role of linguistically motivated input features and pre-processing in advancing dependency parsing of Estonian. In particular, we focus on parsers that take morphological features as explicit inputs, and investigate the effect of a) training data size, b) the choice of lexical and morphological features, and c) clausal syntactic patterns on developing such parsers. While our work indicates that further advancements through a naive increase of training data are hard to obtain, we still confirm the high utility of automatically generated morphological features in parser's input. Our ablation studies indicate that the knowledge about subcategorisation constructions is crucial for parsing, and a targeted search for subcategorisation constructions may lead to more straightforward and effective input features. We also show that decomposing sentences into simpler structures via clausal patterns can lead to performance gains and note that the distribution of clausal subtrees should be considered while increasing training data. pubtype: Academic Journal doctype: Article src: R language: English refInfo: copyright: @attributes: flag: Y custom: Language Resources & Evaluation is a copyright of Springer, 2025. All Rights Reserved. item: Language Resources & Evaluation holder: Springer Nature dt: @attributes: year: 2025 holdings: @attributes: islocal: N |
|---|