JournalFit NLP Data Repository Package

Description

The repository contains processed and analyzed data files generated during the JournalFit-NLP computational evaluation. CSV files provide tabular summaries of dataset composition, retrieval ablation, reranker blending, latency, quantization fidelity, resource monitoring, hardware/software environment, and expanded-pool evaluation metrics. JSON files contain machine-readable metric outputs for baseline, fine-tuned, expanded-pool, and reranking experiments. YAML files document model training and evaluation configurations. PNG and PDF figures visualize the architecture, development flow, retrieval ablations, reranker blending, reranker failure diagnosis, training/thermal monitoring, quantization trade-off, and CPU latency profiles. No private manuscripts, identifiable user records, social media data, or animal data are included.

Steps to reproduce

Download and extract the repository package. Review README.txt, DATA_DESCRIPTION.txt, and file_manifest.csv to understand the dataset structure. Verify file integrity using checksums_sha256.txt. Open the YAML files for experiment settings, JSON files for aggregate metrics, and CSV files for tabular results. The reported tables and figures can be reproduced by loading the CSV/JSON files in Python 3.10+ using pandas and matplotlib. Raw third-party bibliographic metadata and user-uploaded manuscripts are not redistributed because of licensing, privacy, and consent constraints; the repository provides the aggregate metrics, configuration files, figure source data, and supplementary outputs required to reproduce the reported results.

Institutions

• Fırat University

  Elazığ, Elâzığ

Categories

Licence