Update README.md

README.md

@@ -32,9 +32,11 @@ metrics:
 
 ## 📋 Model Summary
 
-This model is a fine-tuned version of climatebert/distilroberta-base-climate-detector. It is specifically engineered to classify shareholder proposals into green (climate/environmental) or non-green categories.
+Shareholder resolutions are often brief and cause ambiguity when considered in isolation. For example, a proposal requesting a report on ``water risk management'' may refer either to environmental water stress (an environmental issue) or to the human right to water access (a social issue). 
 
-Crucially, it is highly effective at isolating environmental topics from **broad, mixed-ESG contexts** without being distracted by generic sustainability or governance buzzwords (etc. Neutrality, Waste, Water...). Trained on a meticulously curated dataset, it achieves an F1 score of 0.981 on the validation set.
+This model is a fine-tuned version of ClimateBERT and specifically engineered to classify shareholder proposals into green (climate/environmental) or non-green categories.
+
+Crucially, it is highly effective at isolating environmental topics from **broad, mixed-ESG contexts** without being distracted by generic sustainability or governance **buzzwords** (etc. Neutrality, Waste, Water...). 
 
 >💡 Designed for: Precision text classification in sustainable finance, ESG analysis, and corporate governance contexts.
 
@@ -78,7 +80,7 @@ dataset = datasets.Dataset.from_dict({"text": [
     """A(An) institutional-type sponsor has filed a shareholder proposal to a(an)
     energy-sector company. This proposal requests: the company to issue a report
     on its greenhouse gas emissions reduction targets.
-    It falls under a broader agenda class: Environmental/Social."""
+    It falls under a broader agenda class: "..."""
 ]})
 
 # ── Inference ──────────────────────────────────────────────────────────────────
@@ -92,7 +94,7 @@ for out in tqdm(pipe(KeyDataset(dataset, "text"), padding=True, truncation=True)
 
 ### 📌 Recommended Input Format
 
-To achieve optimal performance, structure your input text to mirror the training data format:
+To address ambiguity in raw proposal text, we can enhance the model's input with structured proposal- and firm-level context, like the training data format:
 ```
 "A(An) {sponsor_type}-type sponsor has filed a shareholder proposal to a(an)
 {sic2_des}-sector company. This proposal requests: {resolution}.
@@ -104,8 +106,8 @@ relevant to this specific proposal: {AgendaCodeInformation}"
 |:---|:---|:---|
 | `{sponsor_type}` | Type of proposal sponsor | `institutional`, `individual` |
 | `{sic2_des}` | SIC-2 industry sector description | `energy`, `manufacturing` |
-| `{resolution}` | Full text of the proposal resolution | *"the company to report on..."* |
-| `{AgendaCodeInformation}` | ISS agenda code label *(optional but recommended)* | `Environmental/Social` |
+| `{resolution}` | Full text of the proposal resolution | *"Report on Climate Change Performance Metrics Into Executive Compensation Program..."* |
+| `{AgendaCodeInformation}` | Description of ISS agenda code | *"This code is used for proposals seeking..."* |
 
 > 💡 **Tip:** The `{AgendaCodeInformation}` field is optional but including it generally improves prediction confidence, as it provides additional categorical context into brief resolution context.