Create README.md

README.md

@@ -0,0 +1,37 @@
+---
+language: en
+license: apache-2.0
+tags:
+- finance
+- esg
+- sentiment-analysis
+- bert
+metrics:
+- f1
+- accuracy
+---
+
+# finbert_esg_sentiment_classifier
+
+## Overview
+This model is a specialized BERT-based classifier fine-tuned for Environmental, Social, and Governance (ESG) sentiment analysis in financial reports. It categorizes text into specific ESG pillars or identifies neutral financial statements.
+
+
+
+## Model Architecture
+The model utilizes a **BERT-Base-Uncased** backbone with a sequence classification head. 
+- **Encoder**: 12-layer Transformer.
+- **Hidden Dimensions**: 768.
+- **Head**: Linear layer followed by Softmax for 4-class categorization.
+- **Optimization**: Trained using the Cross-Entropy loss function:
+$$\mathcal{L} = -\sum_{c=1}^{M} y_{o,c} \ln(p_{o,c})$$
+
+## Intended Use
+- **Investment Research**: Automating the extraction of ESG signals from 10-K filings and earnings transcripts.
+- **Compliance**: Monitoring corporate communications for ESG-related disclosures.
+- **Sustainable Finance**: Providing data for ESG scoring algorithms.
+
+## Limitations
+- **Context Window**: Restricted to 512 tokens. Long documents must be processed in chunks.
+- **Language**: Optimized for English financial terminology; performance on other languages or casual text is not guaranteed.
+- **Factuality**: Classification is based on linguistic patterns, not external fact-checking of the corporate claims.