Update README.md

README.md

@@ -2,39 +2,22 @@
 license: gpl-3.0
 tags:
 - knowledge-graph
-- rdf
 - wgan
-- gan
+- dblp
 datasets:
 - erdemonal/SemanticGAN-Dataset
 ---
 
-# SemanticGAN
+# SemanticGAN Model Repository
 
-This repository is used as the persistent storage for model checkpoints and processed data produced by the SemanticGAN project.
+This repository contains trained model checkpoints and training logs produced during the SemanticGAN experiments.
 
-Training is performed through a continuous learning pipeline executed via GitHub Actions. 
+The models are trained on the DBLP bibliographic knowledge graph using a Wasserstein Generative Adversarial Network (WGAN). The repository is intended as an archival artifact supporting reproducibility and analysis, rather than as a standalone inference package.
 
-## Project Context
+## Contents
+- gan_latest.pth: latest generator and discriminator checkpoint
+- training_log.csv: recorded loss values across training epochs
 
-SemanticGAN is a research prototype for Knowledge Graph Completion built around a Wasserstein Generative Adversarial Network. The model is trained on the DBLP Computer Science Bibliography and is designed to generate candidate RDF triples that correspond to plausible but previously unobserved relations in the graph.
-
-All source code, training scripts, and technical documentation are maintained in the main GitHub repository at  
-https://github.com/erdemonal/SemanticGAN
-
-A live dashboard showing training metrics and generated outputs is available at  
-https://erdemonal.github.io/SemanticGAN
-
-## Repository Contents
-
-This repository reflects the current state of the training process.
-
-The file `gan_latest.pth` contains the most recent Generator and Discriminator parameters produced by the training pipeline.
-
-The processed dataset `kg_triples_ids.txt` and ID mappings `kg_mappings.json` have been moved to the [SemanticGAN-Dataset](https://huggingface.co/datasets/erdemonal/SemanticGAN-Dataset) repository.
-
-These artifacts are updated automatically after each training run and are consumed by the training and evaluation scripts in the GitHub repository.
-
-## Intended Usage
-
-This repository is not meant to be used for standalone inference. The stored model artifacts are intended to be loaded by the training and evaluation pipeline defined in the main SemanticGAN project, where checkpoint restoration and continued optimization are handled automatically. The pipeline automatically pulls the required data from the companion Dataset repository before starting each training run.
+## References
+- Source code and training pipeline: https://github.com/erdemonal/SemanticGAN
+- Experimental monitoring interface: https://erdemonal.github.io/SemanticGAN