ELTE Poetry Corpus: A Machine Annotated Database of Canonical Hungarian Poetry

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 3471–3478 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 3471 ELTE Poetry Corpus: A Machine Annotated Database of Canonical Hungarian Poetry P´eter Horv´ath1, P´eter Kundr´ath, Bal´azs Indig1, Zs´ofia Fellegi2, Eszter Szl´avich1, T´ımea Borb´ala Bajz´at2, Zs´ofia S´ark¨ozi-Lindner1, Bence Vida1, Aslihan Karabulut1, M´aria Tim´ari1, G´abor Palk´o12 1E¨otv¨os Lor´and University, National Laboratory for Digital Heritage 1088 M´uzeum krt. 6-8., Budapest, Hungary 2Research Centre for the Humanities: Institute for Literary Studies, National Laboratory for Digital Heritage 1118 M´enesi ´ut 11–13., Budapest, Hungary {horvath.peter, indig.balazs, szlavich.eszter, lindner.zsofia, vida.bence, karabulut.aslihan, timari.maria, palko.gabor}@btk.elte.hu, {fellegi.zsofia, bajzat.timea}@abtk.hu, peter.kundrath@gmail.com Abstract ELTE Poetry Corpus is a database that stores canonical Hungarian poetry with automatically generated annotations of the poems’ structural units, grammatical features and sound devices, i.e. rhyme patterns, rhyme pairs, rhythm, alliterations and the main phonological features of words. The corpus has an open access online query tool with several search functions. The paper presents the main stages of the annotation process and the tools used for each stage. The TEI XML format of the different versions of the corpus, each of which contains an increasing number of annotation layers, is presented as well. We have also specified our own XML format for the corpus, slightly different from TEI, in order to make it easier and faster to execute queries on the corpus. We discuss the results of a manual evaluation of the quality of automatic annotation of rhythm, as well as the results of an automatic evaluation of different rule sets used for the automatic annotation of rhyme patterns. Finally, the paper gives an overview of the main functions of the online query tool developed for the corpus. Keywords: poetry corpus, Hungarian, automatic annotation, sound devices 1. Introduction ELTE Poetry Corpus1 is a database with an online query interface2, which stores a significant part of canonical Hungarian poetry with different types of an- notations. Currently, the corpus contains the complete poems of 49 Hungarian authors. The number of po- ems in the corpus is 13,063 and the number of words is roughly 2,7 million. The main format of the cor- pus is TEI XML, which is one of the most widely used formats in the fields of digital humanities and linguis- tics. The annotations of the poems have been created automatically. We annotated the poems on three levels. First, we annotated the structural units of poems: the ti- tles, the stanzas and the lines. Second, we tokenized the texts and annotated the grammatical features of words, i.e. the lemma, the part of speech and the morphosyn- tactic properties. Third, we annotated several features of sound devices as well: the rhyme patterns, the rhyme pairs, the rhythm, the alliterations and the main phono- logical properties of words. In section 2, we give an overview of previous works that served as inspirations for the present project. Sec- tion 3 discusses the stages of the corpus building pro- cess and the tools used for …

Content truncated to 500 words