TL;DR
This paper introduces PaVeDa, a new open-access typological database of verb valency patterns. It significantly expands upon existing resources by incorporating ancient languages, a broader language sample, and linking to external corpora for usage-based examples, enabling richer cross-linguistic and diachronic analysis of verb behavior.
What Problem Does This Paper Address?
Linguistic research has long observed recurring patterns in how verbs take arguments (their "valency") across different languages. To better understand these universal tendencies and variations, typologists have been building databases of verb meanings and their associated argument structures. The existing Valency Patterns Leipzig (ValPaL) database was a significant step, but it has limitations: it lacks ancient languages, has an unbalanced language sample, relies heavily on elicited data rather than real-world usage, and its interface makes cross-linguistic comparison difficult due to language-specific labeling. This makes it challenging to study how verb valency evolves over time or to systematically compare similar patterns across diverse languages.
Methods
The PaVeDa database builds upon the ValPaL framework but introduces several key innovations:
- Expanded Language Sample: PaVeDa aims to include languages from families currently underrepresented in ValPaL (e.g., Uralic, Turkic, Afro-Asiatic) and crucially, ancient Indo-European and Afro-Asiatic languages.
- Diachronic Research Enablement: By including ancient languages, PaVeDa allows for the study of how verb valency patterns have changed over time.
- Usage-Based Data: A major methodological shift is the integration of data from external corpora. This moves beyond elicited examples to provide real-world usage instances of verb patterns, offering a more robust understanding of their attested behavior.
- Cross-Linguistic Annotation Layer: PaVeDa introduces a new layer of annotation for "general" alternations that are not language-specific. This layer maps language-specific patterns to broader conceptual categories, facilitating comparative visualization and analysis across languages.
- Standardized Data Formats: The database adheres to the Cross-Linguistic Data Formats (CLDF) initiative, promoting interoperability and data sharing.
Datasets
The paper describes the PaVeDa database itself as the primary dataset. It is built upon and extends the Valency Patterns Leipzig (ValPaL) database. Additionally, PaVeDa will incorporate data from:
- Ancient Languages: Datasets for Early Latin, Ancient Greek, Gothic, Old Irish, Old English, and Classical Armenian are being extracted or have been extracted.
- Modern Languages: Data will be linked to existing corpora, such as the Russian National Corpus and CORIS (for Italian). For low-resourced languages, data will be supplemented from spoken corpora like KIParla and RadioCast-it.
Key Insights and Ideas
- The Value of Diachrony: Including ancient languages is crucial for understanding the evolution of linguistic structures, revealing how verb valency patterns have developed and changed.
- Corpora as the Gold Standard: Relying on real-world corpus data provides a more accurate and comprehensive picture of verb usage than solely relying on elicited judgments, especially for understanding attested patterns and their frequencies.
- Bridging Language-Specific and Universal Patterns: The new annotation layer for "general" alternations is a clever way to reconcile the need for detailed, language-specific descriptions with the goal of identifying universal linguistic tendencies. This allows for both granular analysis and broad comparison.
- Addressing Data Scarcity: The proposed strategies for documenting low-resourced languages highlight a critical challenge in typological research and offer a practical approach to mitigate it.
Key Results
The paper describes the ongoing construction of the PaVeDa database and its planned features rather than presenting final quantitative results. The key "results" are the architectural and content innovations:
- A More Comprehensive Resource: PaVeDa promises to be a significantly richer resource than ValPaL due to its expanded language coverage (including ancient languages) and its linkage to usage-based data.
- Enhanced Comparative Capabilities: The new annotation layer and interface design are intended to enable simultaneous and contrastive visualization of valency patterns and alternations across languages, overcoming limitations of previous tools.
- Open Access and Interoperability: The database will be freely available and adhere to CLDF standards, promoting collaboration and reproducible research.
Limitations and Future Work
- Ongoing Development: PaVeDa is described as an "ongoing endeavor" and "currently being constructed," meaning its full potential is yet to be realized.
- Low-Resourced Languages: The authors acknowledge the challenge of finding sufficient reference corpora for many languages, both ancient and modern, and propose strategies to address this.
- Interface Refinement: While the new interface aims to improve comparative visualization, the paper implies that further refinement may be needed to fully leverage the data.
- Future Work: The immediate future involves completing the data extraction and integration for ancient and underrepresented languages, refining the annotation system, and making the database publicly accessible. The long-term vision includes applications in first language acquisition and typological language teaching.
Summary updated