Reproducibility: an open-source tool for computational hypothesis testing in natural language
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
We present Reproducibility, an open-source software system for quantitative evaluation of scientific hypotheses formulated in natural language. The system converts scientific statements written in Portuguese or English into executable mathematical functions, assesses their internal consistency through Monte Carlo simulation, and measures their predictive power against real empirical data from open sources (World Bank, NASA POWER, IBGE) or user-supplied CSV files. We propose a composite reproducibility score — combining simulated consistency and empirical fit — and validate it against five cases with expected behaviour established by the literature.
