TL;DR
This paper proposes a novel ontology of systematic linguistic relations to facilitate the creation of shared grammars for related languages, specifically Slavic. By categorizing these relations, the authors aim to reduce the cost and complexity of multilingual grammar engineering, enabling more efficient development and maintenance of computational linguistic resources.
What Problem Does This Paper Address?
Developing computational grammars for multiple languages is a labor-intensive and costly endeavor. While it's intuitive that related languages share linguistic features, effectively leveraging this overlap for grammar engineering has been challenging. Existing approaches often rely on opportunistic "copy-paste-modify" strategies without a systematic linguistic framework. The authors identify a gap in linguistically motivated, structured approaches to grammar sharing, particularly for language families like Slavic, where diverse linguistic theories and formalisms can obscure commonalities. They aim to bridge this gap by defining a conceptual ontology of morphosyntactic relations that can serve as a foundation for shared grammars.
Methods
The core of the paper is the development of a hierarchical ontology of systematic linguistic relations. This ontology is built upon two main dimensions: syntagmatics (how linguistic units combine) and alignment (how they are linearized).
Within syntagmatics, they distinguish:
- Structural syntagmatics: focusing on centricity (endocentric vs. exocentric) and taxis (hypotaxis vs. parataxis), leading to relations like selection and modification.
- Combinatorial syntagmatics: focusing on how these relations are signaled morphologically, including co-variation (agreement) and government.
Alignment is further broken down into continuity, directionality, and periphery. The authors systematically cross-classify these dimensions to derive a comprehensive set of relations, illustrated with examples from Slavic languages. This approach aims to provide a flexible yet structured way to conceptualize and share grammatical knowledge.
Datasets
Not discussed in the paper. The paper focuses on theoretical linguistic analysis and conceptualization, using examples from Slavic languages rather than relying on specific datasets for training or evaluation.
Key Insights and Ideas
- Systematic Relations as the Key to Grammar Sharing: The paper argues that while primitive linguistic entities are often language-specific, the relations between them offer a higher degree of abstraction and are crucial for building shared grammars.
- Ontology as a "Shared Conceptualization": The authors propose using an ontology not as a universal linguistic theory, but as a pragmatic, formal conceptualization of a specific domain (Slavic morphosyntax) to enable consistent grammar design.
- Orthogonal Dimensions for Classification: Distinguishing between syntagmatics and alignment provides a robust framework for analyzing how linguistic units relate to each other and how these relations are expressed.
- Hierarchical Derivation of Relations: The systematic cross-classification of basic dimensions (centricity, taxis, co-variation, etc.) elegantly derives complex linguistic relations like selection, modification, and agreement, providing a principled way to categorize them.
- Slavic Languages as a Testbed: The choice of Slavic languages is motivated by their relatedness, offering a rich ground to explore systematic variations and commonalities in morphosyntax.
Key Results
The paper's primary contribution is the proposed ontology of systematic relations, illustrated through detailed linguistic analysis of Slavic languages. While no quantitative results are presented, the paper demonstrates how this ontology can:
- Provide a unified framework for describing diverse morphosyntactic phenomena across Slavic languages.
- Offer a principled way to identify and leverage shared grammatical structures.
- Facilitate the design of linguistically motivated shared grammars for computational applications.
The paper provides numerous examples (ex. 1-6) from Russian, Bulgarian, and Polish to illustrate how various systematic relations (e.g., relational case, subcategorisation, agreement, co-reference, modification) are realized in these languages, showcasing the ontology's descriptive power.
Limitations and Future Work
Limitations: * The authors acknowledge that their ontology is a "mini-ontology" focused on morphosyntax and does not aim to be an all-encompassing theory of language. * The paper primarily focuses on theoretical conceptualization and linguistic analysis, with no empirical evaluation of its impact on actual grammar engineering efficiency or performance. * The distinction between certain relations, particularly in exocentric cases, can still be theory-dependent and may require further refinement.
Future Work: * The authors suggest that grammars designed using this ontology can be extended and employed in HPSG-accommodating systems, implying future work in computational implementation. * The insights gained from this comparative linguistic description are expected to benefit psycholinguistic research on bilingualism and second language acquisition. * The approach is also seen as useful for CALL applications and for computational treatment of cross-language interference in grammar and style checking. * Further exploration of the systematic relations of parallelism and coordination is mentioned as a direction for future work.
Summary updated