Itzulpengintza Automatikoa

Machine translation (MT) has benefited from using synthetic training data originating from translating monolingual corpora, a technique known as backtranslation. Combining backtranslated data from different sources has led to better results than when using such data in isolation. In this work we analyse the impact that data translated with rule-based, phrase-based statistical and neural MT systems has on new MT systems.

This article describes a compositional distributional method to generate contextualized senses of words and identify their appropriate translations in the target language using monolingual corpora. Word translation is modeled in the same way as contextualization of word meaning, but in a bilingual vector space. The contextualization of meaning is carried out by means of distributional composition within a structured vector space with syntactic dependencies, and the bilingual space is created by means of transfer rules and a bilingual dictionary.

This is a summary of the PhD thesis written by Uxoa Iñurrieta under the supervision of Dr. Gorka Labaka and Dr. Itziar Aduriz. Full title of the PhD thesis in Basque: "Izena+aditza Unitate Fraseologikoak gaztelaniatik euskarara: azterketa eta tratamendu konputazionala". The defense was held in San Sebastian on November 29, 2019. The doctoral committee was integrated by Ricardo Etxepare (Centre National de la Recherche Scientifique), Margarita Alonso (Universidad de Coruña) and Miren Azkarate (University of the Basque Country).

Multiword Expressions (MWEs) are combinations of words which exhibit some kind of idiosyncrasy. Due to their idiosyncratic nature, they pose several problems to Natural Language Processing (NLP). In this PhD, two of the most challenging tasks concerning MWE processing are addressed: the automatic identification of MWE occurrences in corpora and their translation in Machine Translation (MT).

Unitate Fraseologikoak (UFak) hizkuntzek bere-bereak dituzten hitz-konbinazio idiomatikoak dira. Hizkuntzaren Prozesamenduko (HPko) tresnek kalitatezko emaitzak izan ditzaten, beharrezkoa da halakoak ondo tratatzea, baina lan horrek hainbat zailtasun ditu; besteak beste, hitzez hitzeko itzulgarritasun eza. Tesi-lan honetan, aditza+izena motako UFen azterketa linguistiko bat egin dugu, halakoek HPren alorrean sortzen dituzten bi arazo garrantzitsuri aurre egiten laguntzeko: batetik, corpusetan UFak automatikoki identifikatzeari, eta bestetik, UF horiek gaztelaniaren eta euskararen