Building the Largest Open-Access Multilayer Corpus for Ancient Greek: Opera Graeca Adnotata

Core Concepts

The Opera Graeca Adnotata (OGA) is the largest open-access multilayer corpus for Ancient Greek, containing 1,687 literary works and over 34 million tokens, with seven annotation layers including tokenization, sentence segmentation, lemmatization, morphology, dependency, dependency function, and Canonical Text Services (CTS) citation.

Abstract

The Opera Graeca Adnotata (OGA) is a multilayer corpus for Ancient Greek (AG) that presents the beta version 0.1.0 of the largest open-access corpus for this language. The corpus consists of 1,687 literary works and 34,172,140 tokens, extracted from the PerseusDL and OpenGreekAndLatin GitHub repositories.
The texts have been enriched with seven annotation layers:

Tokenization layer: A rule-based algorithm is used to tokenize the texts, which are extracted from complex EpiDoc TEI XML files.
Sentence segmentation layer: A rule-based algorithm is used to identify sentence boundaries based on punctuation marks.
Lemmatization layer: The lemmas are provided by the COMBO parser, which was trained on the Ancient Greek Dependency Treebank (AGDT) data.
Morphological layer: The COMBO parser also provides the morphological annotation, following the AGDT schema.
Dependency layer: The COMBO parser outputs the dependency relations between tokens.
Dependency function layer: The COMBO parser also provides the dependency function labels.
Canonical Text Services (CTS) citation layer: The CTS citation for each token is extracted from the EpiDoc TEI XML files.

The corpus is released in the standoff formats PAULA XML and LAULA XML to ensure scalability and reusability. PAULA XML is used as a serialization format for the ANNIS query engine, while LAULA XML is a more efficient XML structure for parsing the texts.

Stats

The corpus contains 34,172,140 tokens from 1,687 literary works.

Quotes

No notable quotes were extracted from the content.

Key Insights Distilled From

Opera Graeca Adnotata

by Giuseppe G. ... at arxiv.org 04-02-2024

https://arxiv.org/pdf/2404.00739.pdf

Deeper Inquiries

How can the multilayer annotation of OGA be leveraged for advanced linguistic analysis of Ancient Greek texts?

The multilayer annotation of OGA provides a rich dataset that can be utilized for in-depth linguistic analysis of Ancient Greek texts. Researchers can leverage the tokenization layer to study word frequencies, patterns, and variations in usage across different literary works. The sentence segmentation layer allows for the analysis of sentence structures, syntactic patterns, and discourse organization within the texts. The lemmatization layer provides insights into the root forms of words, enabling researchers to explore lexical relationships, word derivations, and semantic shifts over time. The morphological layer offers detailed information on the grammatical features of words, facilitating morphosyntactic analysis and studies on language evolution. The dependency and dependency function layers allow for the investigation of syntactic relationships, sentence parsing, and the identification of grammatical roles within sentences. Finally, the CTS citation layer enables scholars to reference specific passages accurately, facilitating intertextual analysis, source comparison, and textual criticism.

What are the potential limitations or biases in the text selection and annotation process of OGA, and how could they be addressed in future versions?

One potential limitation in the text selection process of OGA is the reliance on specific repositories like PerseusDL and OpenGreekAndLatin, which may introduce biases towards certain types of texts or authors. To address this, future versions of OGA could consider expanding the sources to include a more diverse range of Ancient Greek texts from various genres, time periods, and authors. Additionally, the automatic encoding normalization applied to address character inconsistencies may introduce errors or inaccuracies, especially in cases where manual intervention is required. Future versions could implement more sophisticated algorithms or manual verification processes to ensure accurate encoding normalization. In terms of annotation, the rule-based algorithms used for tokenization, sentence segmentation, and morphosyntactic annotation may not capture all linguistic nuances or variations present in the texts. To mitigate this, future versions could incorporate machine learning models trained on a larger and more diverse dataset to improve the accuracy and coverage of the annotations.

What other types of linguistic annotation (e.g., semantic, pragmatic, prosodic) could be added to OGA to further enhance its utility for scholars of Ancient Greek?

In addition to the existing annotation layers, adding semantic annotation to OGA would enhance its utility for scholars of Ancient Greek by providing information on the meaning and interpretation of words and phrases within the texts. Semantic annotation could include the identification of semantic roles, named entities, and semantic relationships between words. Pragmatic annotation could focus on the contextual use of language, implicatures, speech acts, and discourse markers, offering insights into the pragmatic aspects of communication in Ancient Greek texts. Prosodic annotation, which captures intonation, stress patterns, and rhythm in speech, could provide valuable information on the oral performance of the texts and aid in understanding the poetic and rhetorical elements present in Ancient Greek literature. By incorporating these additional layers of linguistic annotation, OGA would offer a more comprehensive and nuanced analysis of Ancient Greek texts, catering to a wider range of research interests and methodologies.