A Database for Modal Semantic Typology

Qingxia Guo, Nathaniel Imel, Shane Steinert-ThrelkeldNAACL 2022

TL;DR

This paper introduces a new, publicly accessible database designed to consolidate and standardize cross-linguistic data on modal semantics. By providing a structured repository of how different languages express concepts like possibility and necessity, the database aims to facilitate empirical verification of semantic universals and drive new theoretical investigations into the nature of modality.

What Problem Does This Paper Address?

Modality – the linguistic expression of concepts like possibility, necessity, obligation, and permission – is a fundamental aspect of human language, yet its cross-linguistic variation is complex and has been challenging to systematically study. While linguists have gathered significant data on how different languages express modality, this knowledge is often scattered across various descriptive grammars and specialized studies, making it difficult to compare languages or test broad theoretical claims. A key challenge is the lack of a standardized, theory-neutral framework for representing modal semantics, particularly the interplay of "force" (e.g., necessity vs. possibility) and "flavor" (e.g., epistemic vs. deontic). This paper addresses this gap by proposing and releasing a database that captures this variation in a structured and accessible format.

Methods

The core of the paper is the introduction of a novel database schema and structure for representing modal semantic typology. The authors propose a theory-neutral approach that focuses on identifying the specific pairs of "force" and "flavor" that a given modal expression in a language can convey.

  • Data Schema: The database is structured around observations, where each observation represents a judgment about whether a particular modal expression in a specific language can express a given (force, flavor) pair. This is captured in a "basic format" using a table with columns for expression, force, flavor, and can_express (a binary indicator).
  • Theory Neutrality: The framework aims to be flexible, allowing for a wide range of force and flavor values without committing to a specific theoretical interpretation of their underlying semantics. This allows for the inclusion of data from diverse linguistic traditions and theoretical perspectives.
  • CLDF Format: To enhance interoperability and reusability, the database is also made available in the Cross-Linguistic Data Formats (CLDF) standard. This format facilitates integration with other linguistic datasets and enables more sophisticated querying and analysis.

Datasets

The paper introduces the Modal Typology Database itself as the primary dataset.

  • Content: The database consolidates existing typological knowledge on modal semantics from published sources, including descriptive grammars and specialized fieldwork.
  • Scope: At the time of writing, the database contains data from 17 languages, with a mix of data derived from detailed semantic fieldwork (considered more robust due to explicit negative judgments) and descriptive grammars.
  • Structure: The data is organized by language, with individual files for metadata (metadata.yml) and modal expressions (modals.csv). Aggregated files are also provided for easier access.

Key Insights and Ideas

  • Standardization is Crucial: The paper highlights the critical need for standardized data formats and theory-neutral frameworks to advance cross-linguistic semantic research, particularly in complex domains like modality.
  • Force and Flavor as Core Axes: The authors reinforce the established linguistic insight that modal semantics can be effectively analyzed along the axes of "force" and "flavor," and propose a practical way to catalog these variations.
  • Variable Force/Flavor Modals are Common: The database aims to capture languages where modals exhibit variability in force or flavor, a phenomenon observed in languages like St'át'imcets and Washo, which challenges simpler, fixed-meaning accounts of modals.
  • Empirical Verification of Universals: The database is designed to be a powerful tool for testing proposed semantic universals, such as the "Independence of Force and Flavor" (IFF) universal, by allowing researchers to search for counterexamples.

Key Results

The paper presents the Modal Typology Database as a resource. While not reporting on a specific empirical study using the database to discover new findings, it demonstrates its utility:

  • Database Availability: The database is publicly available at https://clmbr.shane.st/modal-typology.
  • Data Snapshot: As of the paper's writing, the database contains data from 17 languages and 435 unique observations (judgments about modal expression capabilities).
  • IFF Universal Check: The authors demonstrate how the database can be used to verify semantic universals. They provide a Python script that checks for violations of the IFF universal, noting that at the time of writing, no counterexamples were found in the database.

Limitations and Future Work

  • Data Completeness: The authors acknowledge that data from descriptive grammars often lack explicit negative judgments (i.e., what a modal cannot do), which can limit the robustness of analyses. Data from detailed semantic fieldwork is preferred for this reason.
  • Ontology of Force/Flavor: While aiming for theory neutrality, the authors note that the list of forces and flavors may not be exhaustive or perfectly disjoint, requiring potential future refinement and mapping.
  • Graded Modality: The current schema primarily focuses on weak, strong, and weak necessity modals, setting aside the full range of graded modality (e.g., probabilistic expressions) due to a lack of consensus on their typology. However, they suggest the schema could be extended to accommodate this.
  • Future Work: The authors plan to expand the database with more data, build visualization tools for easier interaction, and potentially extend the data schema to include syntactic information about modal expressions. They also suggest exploring crowdsourcing methods to gather more comprehensive data, especially for under-studied languages.

Summary updated

BibTeX
@inproceedings{guo2022database,
  author = {Qingxia Guo and Nathaniel Imel and Shane Steinert-Threlkeld},
  title = {{A Database for Modal Semantic Typology}},
  booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2022},
}
Figure