PaperlayerView exact revision on arXiv

Source-linked AI summary

Translations in library catalogues: how to find them?

Péter Király

arXiv:2610.32176v1cs.DL

TL;DR

Research using library records often relies on derivative data, and extracting translations from raw MARC records is not straightforward. This paper describes MARC21 translation cues and applies them across five national-library catalogues. The analysis finds useful translation information but uneven coverage and incomplete records.

  • Problem

    Researchers rarely work directly with raw MARC records, and no comprehensive account had described extracting translations from them.

    Source: Introduction · p. 1

  • Method

    The paper analyzes MARC21 elements for detecting translations and extracting related information, then applies the process to five national-library catalogues.

    Source: Introduction · p. 1, Data elements · p. 3

  • Results

    Translator information appears in 63.5–73.5% of records except BN, while original-title information appears in 7–34% except KB; overall extracted data are precise but incomplete.

    Source: Evaluation · pp. 6–7

  • Takeaways & Limitations

    Catalogue records provide substantial information about translations, but researchers should supplement them with other sources when completeness matters.

    Source: Evaluation · p. 7

    Detection results show limitations and biases, and the study notes that examined catalogues rarely used the 765 field that could record original-work details.

    Source: Data elements · p. 3, Further plans · p. 8

Abstract

from arXiv · show

In library catalogues based on MARC21 bibliographic metadata schema there are about a dozen data elements that catalogers can utilize to record information about the different aspects of translations: whether or not the edition is a translation, source and target language, title of the original work, translator and other contributors, publication data of the original work, literary genre. Libraries however rarely explore the full arsenal, thus to find translation is not a trivial task. The paper describes these important MARC21 data elements, and shows a use case of finding translations in five catalogues (of the Norwegian, Croatian, Polish, Belgian and Dutch national libraries) explaining the differences among the practices, and the results.

Introduction

The paper addresses the limited use of raw MARC21 records in quantitative translation and book-history research, and the lack of a comprehensive account of extracting translations from them. It describes relevant MARC21 elements, harmonisation steps, and translation findings across five national-library catalogues.

Introduction

  • Quantitative studies of translation and book history rarely use raw MARC21, UNIMARC, or PICA records, and prior work has not comprehensively described extracting translations from raw MARC data.Researchers often rely on derivative library formats, catalogue interfaces, or extraction services.Source: Introduction · p. 1
  • The paper presents relevant MARC21 elements, data-harmonisation steps, and translation-detection results from five national-library catalogues.It also reports basic patterns of Hungarian translations.Source: Introduction · p. 1
  • The study uses QA Catalogue to analyze translation information in library records.The paper describes a use case across Norwegian, Croatian, Polish, Belgian, and Dutch national libraries.Source: Introduction · p. 1

Data elements

Translation detection requires combining MARC21 elements because catalogue records and practices are incomplete and inconsistent. The paper describes cues for identifying translations and extracting their related languages, titles, contributors, and original-publication details.

Data elements

  • Catalogue incompleteness, changing schemas, and mixed cataloguing practices mean 041/ind1 alone rarely identifies translations reliably.The paper therefore recommends checking several elements that reflect different aspects of translation.Source: Data elements · pp. 1–2
  • Initial cues include item and text language codes, the 041 translation indicator and original-language code, translated-by wording in 245$c, and a uniform title in 240$a.The paper treats these as an initial field set that may be extended as more recording practices are learned.Source: Data elements · p. 2
  • Translator entries in 700$e or 700$4 and general notes in 500$a can provide evidence, but varied and ambiguous phrases require review and catalogue-specific phrase handling.General notes may refer to languages or contributors and have no strict content rules; terms such as Hungarian átdolgozta can refer to several transformative actions.Source: Data elements · p. 3
  • The uniform title in 240$a often gives a translated work’s title, but the field also serves other uses; 240$l denotes the current entry’s language.The paper notes that genre information can help distinguish uses of uniform titles.Source: Data elements · p. 3
  • The 765 original-language entry can record original title and publication details, but the examined catalogues used it very rarely.Its attributes include the original-language entry’s title, uniform title, and place, publisher, and date of publication.Source: Data elements · p. 3

Data harmonisation

The paper harmonises extracted language, date, place, and personal-name information to support translation analysis. It uses controlled language codes, pattern-based date processing, and a dedicated place-name repository, while leaving personal-name normalisation outside the study.

Data harmonisation

  • MARC language codes use a Library of Congress-controlled vocabulary, making language information more standardised than publication dates and places.The paper identifies source and target language as extraction targets.Source: Data harmonisation · p. 4
  • Publication dates appear in varied forms, so QA Catalogue uses regular expressions for common patterns and reports unhandled forms with recognition statistics.Examples include bracketed corrections, Roman numerals, textual additions, and dates from other calendar systems.Source: Data harmonisation · p. 4
  • A dedicated repository maps frequent publication-place variants and synonyms to standard names, including multiple-place records, and reports unknown forms.The repository stores standardised names, Geoname identifiers, countries, coordinates, and common library notations.Source: Data harmonisation · p. 4
  • Personal-name normalisation using services such as VIAF, ISNI, or Wikidata is suggested, but was outside this study’s focus.Source: Data harmonisation · p. 4

Evaluation

The study applies translation detection to five national-library catalogues and finds that catalogue practices vary substantially in which translation-related MARC21 fields they populate. Records often support identification of translators and target languages, but provide incomplete information about original titles and publication details, so researchers may need other sources.

Evaluation

  • The evaluation covered five national libraries: Norway (2.28 million records), Croatia (1.65 million), Poland (584,000 literary works), Belgium (4.34 million) and the Netherlands (5.38 million).Detected translation shares were 17.0%, 12.8%, 29%, 4.51% and 14%, respectively; Poland's analysis was limited to literary works.Source: Evaluation · p. 5
  • Catalogue practices differ: translation indicators, original-language codes, translator identification and original-work fields have uneven coverage across the five catalogues.The 041/ind1 translation indication is very low at KBR, where around 80% of records say no information was provided; original-work information is rarely recorded except in KB.Source: Evaluation · p. 6
  • Translator information appears in 63.5–73.5% of catalogues’ detected translations, except BN; original titles appear in only 7–34%, except KB.Researchers get almost no information about original publication details, while source-language coverage varies by catalogue and target-language coverage is generally high except in NB.Source: Evaluation · p. 6
  • Comparisons of an author’s translation patterns across regions require sources beyond catalogue records.The extracted dataset has high precision but incomplete data-element coverage; the paper suggests supplementing records with other catalogues, bibliographies, and other sources, or using detected authors as a starting point for further searches.Source: Evaluation · p. 7

Patterns of translation23

The paper uses detected Hungarian translations to illustrate differing patterns across national catalogues. The examples reveal distinct author, language, and publication-place profiles, while the Norwegian results are constrained by low target-language coverage and the lack of duplicate detection.

Patterns of translation23

  • Norwegian results are biased by extremely low target-language coverage; the records also may count multiple editions or duplicate copies because the method lacks duplicate detection.Only three authors have multiple records with both source and target language, and none of Imre Kertész’s nine records are detected as translations.Source: Patterns of translation23 · p. 7
  • Croatian catalogue records include translations into several former Yugoslav languages as well as Croatian, with differing patterns across languages.The leading Hungarian authors translated into Croatian include Ferenc Molnár (29 records), Sándor Márai (27), Éva Janikovszky (19), Béla Hamvas (18) and Imre Kertész (18).Source: Patterns of translation23 · p. 7
  • In the Polish catalogue, Sándor Márai leads Hungarian authors with 88 translated records, followed by Ferenc Molnár with 32.Most were published in Warsaw; the catalogue also includes less prominent Hungarian authors and a comic-screenplay writer, Tibor Cs. Horváth.Source: Patterns of translation23 · p. 8
  • Dutch records show Márai translations dating to 1934, earlier than the Croatian and Polish translations, which began in 2000.Amsterdam is the main publication place, though Hungarian translations also appeared in nearly 40 other cities in the Netherlands or Flemish Belgium.Source: Patterns of translation23 · p. 8

Further plans

The authors identify limitations and biases in automatic translation detection and outline improvements to the method and its evaluation.

Further plans

  • The current detection process highlights translation patterns but also has limitations and biases.The authors propose expanding linguistic patterns, normalizing personal names, adding ignored MARC fields, and accounting for library-specific cataloguing practices and duplicates.Source: Further plans · p. 8
  • Future evaluation should compare model results against a manually selected gold-standard set of translations.The proposed evaluation would calculate precision, recall and F1; other planned steps include authority files and collaboration with researchers and librarians.Source: Further plans · p. 8

Literature

The literature list includes work on MARC21 and interpreting bibliographic records, research on translation identification and source languages, and quantitative translation history.

Literature

  • The cited bibliography includes work on MARC21 as data and on interpreting bibliographic records.It lists Karen Coyle’s “MARC21 as Data: A Start” and Jason Thomale’s “Interpreting MARC: Where’s the Bibliographic Data?”Source: Literature · p. 9
  • The references include studies of identifying translations and source languages in national bibliography and catalogue records.Examples include Laura Ivaska’s study of the Finnish National Bibliography and Lisa Maria Teichmann’s thesis on German fiction in translation in the German National Library catalogue.Source: Literature · p. 9
  • The references include resources on validating library data and on MARC21 bibliographic metadata.They include Shacl4Bib, the QA Catalogue tool, and the Library of Congress introduction to MARC 21 Bibliographic.Source: Literature · p. 9
  • The bibliography also points to bibliographic-data research and quantitative translation history.It lists Zhou and Sun’s work on bibliography-based quantitative translation history and Király’s bibliography of bibliographic data science.Source: Literature · p. 9
Loading 2610.32176v1…