CODE / RESEARCH INFRASTRUCTURE
FOSP
Figures of Speech Pipeline
A computational pipeline for detecting, classifying, reconstructing, and studying figurative language across large scholarly corpora.
- v7
- Pipeline
- Python
- Runtime
- ML + LLMs
- Methods
- 883
- EJIL book corpus
Architecture
- 01
Detection
- 02
Clustering
- 03
Argument mapping
- 04
Typology
- 05
Classification
- 06
Analytics
What it does
FOSP (Figures of Speech Pipeline) is research infrastructure for studying figurative language in scholarly writing at corpus scale. It is built to run over large document collections — not as a one-off notebook experiment.
The pipeline detects candidate figures, clusters related usages, maps them onto argumentative roles, builds typologies, classifies instances, and produces analytics for close reading and book-level claims.
Architecture
Current production line is pipeline v7 (Python runtime). Stages are listed above. Methods combine classical corpus techniques, machine learning, and LLM-assisted reconstruction, with evaluation protocols running alongside production runs.
Corpus numbers
883 is the EJIL article count used as the flagship book corpus for Figuring International Law (corpus span noted on that project page).
Broader internal research collections processed during pipeline development are not presented here as the empirical study denominator. When a larger public corpus figure is released, it will be stated with its defined scope.
Validation
Reliability and demotion rules for LLM-as-coder measurement are under development in methods papers that remain unfinished (human validation pending). Those drafts stay under this project until they are finished manuscripts.
Public status
No public GitHub repository is linked yet. When a flagship repository is released, this page will carry the link, schemas, methodology notes, and sample outputs.