UD_Irish-Cadhan

Treebank card from UD release 2.18: metadata and texts come from UD_Irish-Cadhan/README.md; size is counted by streaming the .conllu files and checked against the release stats.xml. Generated by tools/treebanks build; do not edit by hand. Papers: Papers: UD_Irish-Cadhan; seeds: seeds/; converter: convert.md.

Annotation layers

UD scale: manual native — annotated by hand directly in UD; converted from manual — converted from manual annotation in another scheme; converted with corrections — converted and partly corrected by hand; automatic with corrections — automatic, with partial manual corrections; automatic — automatic; not available — the layer is absent.

Size

Tokens are surface tokens (Spanish vámonos is one), words are syntactic words (vamos + nos); multiword tokens are those split into several words; empty nodes are the N.M nodes of the enhanced graph. Splits train, dev are absent from the release.

Summary

This is the Cadhan Aonair UD treebank, consisting of 150 sentences randomly sampled from six pre-standard Irish texts. It was subsequently augmented with a late Early Modern Irish syllabic poem representing 43 sentences, described in a separate section below.

Introduction

Irish underwent a major spelling standardization in the 1940’s and 1950’s, and as a result it can be challenging to apply modern language technologies to older, “pre-standard” texts. For many years now, the general strategy for tagging and parsing older Irish texts has been to pre-process them with an automatic standardizer (Scannell, 2014), and to then use existing tools designed for the modern language. This approach has been successful, but has some inherent limitations. First and foremost, since there are no resources for directly tagging or parsing pre-standard texts, the standardizer must do its job without the benefit of linguistic annotations. This places an upper bound on the performance of the standardizer, and therefore on the full pipeline for analyzing older texts. In addition, there are certain grammatical phenomena that have all but disappeared in the modern language (e.g. the dative case); these cannot be properly handled with the existing approach.

Our primary aim in creating this treebank was to establish a test set for evaluating lemmatization, tagging, and parsing of pre-standard Irish texts. This should enable experimentation with various approaches that we hope will eventually outperform the existing pipeline. Although the test set is quite small (150 sentences, 3804 tokens), we hope to expand it enough to allow the training of a parser designed to act directly on pre-standard texts.

The corpus contains 25 sentences each from six different books published between 1602 and 1936. Texts published in the late 19th century and early 20th century are much easier to process than older texts. The orthography, while quite different from the standard, is much more consistent than what one finds in texts published before the 1880s. We selected three books published in this later period, one from each of the major Irish dialects: Deoraidheacht by Pádraic Ó Conaire (1910, Connacht Irish), Peig by Peig Sayers (1936, Munster Irish), and Scairt an Dúthchais, a translation of Jack London’s Call of the Wild by Niall Ó Domhnaill (1932, Ulster Irish). We then selected three older (and consequently more challenging) texts to round out the corpus: Foras Feasa ar Éirinn by Seathrún Céitinn (1634), the 1602 translation of the Gospel of John by Uilliam Ó Domhnaill, and Cín Lae Amhlaoibh, a diary kept by Amhlaoibh Ó Súilleabháin between 1827 and 1835.

The annotations were produced by standardizing the texts, parsing them with a UDPipe model trained on the modern Irish treebank, projecting the annotations back to the source texts, and then manually correcting the results. Full details are available in Scannell (2022).

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/treebanks/ga/cadhan/