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: $$\mathcal{L}=-\sum _{c=1}^M{y}_{o,c}\ln (p_{o,c})\backslash mathcal\{L\}\; =\; -\backslash sum\_\{c=1\}^\{M\}\; y\_\{o,c\}\; \backslash ln(p\_\{o,c\})$$

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.