A Frame-Semantic Parsing Plugin for Swedish Research Infrastructure

Abstract

We present the development of a frame-semantic parsing plugin for Sparv – an annotation pipeline for Swedish. The plugin integrates a frame-semantic parser into the annotation pipeline, enabling the automatic identification of frames evoked by lexical units and the assignment of frame elements to their corresponding arguments in text. Designed to operate seamlessly within the infrastructure, the plugin takes raw text as input, and outputs semantic role information in a standardized format compatible with other annotation layers. This implementation demonstrates how frame-semantic analysis can be made available as an additional corpus annotation layer, enriching texts with structured semantic representations that go beyond syntactic or lexical features. By providing access to semantic role information, the plugin can support a wide range of research applications, including semantic search, discourse analysis, and investigation of meaning variation in language use.