FrameNetin

FrameNetin is a multilingual frame-based lexical semantic resource proposed to extend Frame Semantics into modern natural language processing. It organizes meaning around semantic frames—schemas that describe participants and relations in typical situations—and links them to lexical units in multiple languages.

The project provides a curated set of frames, a cross-linguistic frame alignment map, inventories of frame elements, and annotated lexical units. It supports dynamic frame creation to accommodate neologisms and domain-specific terminology, with guidelines to ensure consistency across languages.

FrameNetin builds on established FrameNet methods, collecting data from corpora, expert annotation, and semi-automatic extraction. Annotations attach lexical units to frames and label frame elements with semantic roles, with example sentences illustrating usage. Data are stored in interoperable formats and accompanied by licensing and provenance information.

Access and tools: FrameNetin offers downloadable datasets, an API for querying frames and mappings, and an annotation interface for contributors. It is intended for use in research and industry NLP tasks, including semantic role labeling, machine translation, and cross-lingual information extraction.

Limitations include coverage gaps in less-resourced languages, potential ambiguity in frame-to-lemma mappings, and the need for ongoing maintenance as language evolves. Future work focuses on expanding multilingual coverage, improving alignment with related resources, and integrating FrameNetin with neural NLP pipelines.