A Linked Coptic Dictionary Online

Frank Feder, Maxim Kupreyev, Emma Manning, Caroline T. Schroeder, Amir ZeldesCOLING 2018

TL;DR

This paper introduces the "Coptic Dictionary Online" (CDO), a freely accessible, web-based dictionary for the Coptic language. It addresses the lack of integrated digital resources for Coptic by linking dictionary entries to corpora, providing multilingual definitions, and leveraging NLP tools for enhanced searchability and frequency analysis.

What Problem Does This Paper Address?

Coptic, the final stage of the ancient Egyptian language, is a crucial link in understanding Egypt's linguistic history. Despite its importance, digital resources for Coptic are scarce, particularly comprehensive, freely available dictionaries. Existing lexicographical work often overlooks loanwords and lacks integration with textual corpora, hindering linguistic research. This paper tackles the challenge of creating a unified, accessible digital resource that bridges this gap, making Coptic vocabulary and its usage more discoverable and researchable.

Methods

The CDO project employs several key methods:

  • TEI XML Encoding: Dictionary data is encoded using the Text Encoding Initiative (TEI) XML standard, specifically its Dictionary module, ensuring structured and interoperable data.
  • Cross-referencing and Linking: Entries are extensively cross-referenced, linking to Crum's Coptic Dictionary, other dictionaries, and crucially, to the Coptic Scriptorium corpora. This creates a rich network of linguistic information.
  • Multilingual Definitions: Definitions are provided in English, French, and German, increasing accessibility for a wider range of researchers.
  • NLP Integration: Natural Language Processing (NLP) tools are used to handle inflected forms, enabling users to search for words even if they are not in their base dictionary form. This includes automatic lemmatization for look-up and corpus analysis.
  • REST API: A REST API allows external projects to access dictionary headwords, facilitating integration with other digital humanities tools.
  • Web Interface: A user-friendly web interface allows for sophisticated searches based on Coptic word forms, dialects, parts of speech, and definitions, with options for exact matches or keyword searches.

Datasets

  • Coptic Dictionary Online (CDO) Dataset: This dataset is primarily based on W. Crum's "Coptic Dictionary" (2000 [1939]), containing 8,042 entries and 18,150 word forms. It is being extended with data from other Coptic dictionaries and aims to integrate loanwords.
  • Coptic Scriptorium Corpora: These publicly available corpora, containing over 500,000 tokens from numerous Coptic documents, are used for evaluation and provide the basis for corpus frequency information and linked searches.
  • Thesaurus Linguae Aegyptiae (TLA): A foundational dataset for the CDO, providing a digital offspring of a long-term Egyptian language research project.
  • Database and Dictionary of Greek Loanwords in Coptic (DDGLC): Data from this project is being integrated to address the lack of loanword coverage.

Key Insights and Ideas

  • The Power of Linked Data: The paper strongly advocates for the "linked data" approach, demonstrating how connecting dictionary entries to corpora and other linguistic resources dramatically enhances their utility and research potential.
  • Bridging the Lexicon-Corpus Divide: A significant contribution is the successful integration of a dictionary with large-scale corpora, enabling frequency analysis and contextualized word look-up, which was previously a major gap.
  • Handling Morphological Complexity: Coptic's agglutinative nature and complex morphology (e.g., incorporated verbs, bound groups) present unique challenges for lexicography and NLP. The CDO's approach to encoding and searching these forms is a key innovation.
  • Addressing Loanword Neglect: The paper highlights the historical neglect of loanwords in Coptic lexicography and outlines a clear strategy for their integration, recognizing their importance in understanding the language's evolution.
  • User-Centric Design: The development of a flexible web interface with multiple search options and the use of NLP for inflected forms demonstrate a commitment to making the resource accessible and useful for a broad audience.

Key Results

  • The CDO dataset is based on Crum's dictionary, covering a substantial portion of native Coptic vocabulary.
  • Evaluation on the Coptic Scriptorium corpora shows high token coverage (over 97% for native common nouns when ignoring loanwords and proper names).
  • The NLP pipeline achieves a lemmatization accuracy of 97.23% on correctly segmented text.
  • The web interface allows users to search by Coptic word form, dialect, part of speech, and definition.

Limitations and Future Work

  • Loanwords: The dictionary currently has limited coverage of loanwords (Greek and Arabic), although integration is underway.
  • Dialectal Forms: While some dialectal forms are included, the focus is primarily on Sahidic, the classical dialect.
  • Inflected Forms: Inflected forms of nouns are not directly included, though auto-redirection from inflected forms is supported.
  • Multi-word Expressions: While linked search for multi-word expressions is supported, a reverse look-up from corpora to multi-word dictionary entries is not yet implemented.
  • Future Work: The authors plan to integrate loanword definitions, non-Sahidic vocabulary, and links for multi-word entries from corpora. They also aim to explore fuzzy search for unknown inflected forms and build frequency visualizations.

Summary updated

BibTeX
@inproceedings{feder2018linked,
  author = {Frank Feder and Maxim Kupreyev and Emma Manning and Caroline T. Schroeder and Amir Zeldes},
  title = {{A Linked Coptic Dictionary Online}},
  booktitle = {International Conference on Computational Linguistics},
  year = {2018},
}
Figure