GE-Lab
/

SDGs-classifier

Text Classification

multi-label-classification

Model card Files Files and versions

GE-Lab commited on Aug 5, 2025

Commit

5cf7060

·

verified ·

1 Parent(s): 4dd4801

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -132,7 +132,7 @@ for i, prob in enumerate(probabilities):
 ### Training Data
 The model was trained on a novel, heterogeneous corpus of 23,969 multi-labeled documents from 11 diverse sources, including government, academia, industry, and civil society, with some sources translated from Japanese. This approach was designed to address the "interpretive diversity" of SDG-related language.
-For full details on reconstructing the training corpus, please refer to **Supplementary Information S4** in our paper.
 ### Evaluation
 This model was selected based on its superior generalization performance (especially recall) on external datasets like the OSDG Community Dataset and the SDGi Corpus. On a human-coded sample of scientific articles, the model achieved a macro-averaged **F1-score of 0.623**. For a full breakdown of performance metrics, please see the paper.

 ### Training Data
 The model was trained on a novel, heterogeneous corpus of 23,969 multi-labeled documents from 11 diverse sources, including government, academia, industry, and civil society, with some sources translated from Japanese. This approach was designed to address the "interpretive diversity" of SDG-related language.
+For full details on reconstructing the training corpus, please refer to **Supplementary Information TableS1** in our paper.
 ### Evaluation
 This model was selected based on its superior generalization performance (especially recall) on external datasets like the OSDG Community Dataset and the SDGi Corpus. On a human-coded sample of scientific articles, the model achieved a macro-averaged **F1-score of 0.623**. For a full breakdown of performance metrics, please see the paper.