Add metadata for license, library, and pipeline tag and add paper/code links

README.md

@@ -1,18 +1,25 @@
 ---
+base_model: ExponentialScience/LedgerBERT
 datasets:
 - ExponentialScience/DLT-Sentiment-News
 language:
 - en
-base_model:
-- ExponentialScience/LedgerBERT
+library_name: transformers
+license: cc-by-nc-4.0
+pipeline_tag: text-classification
 ---
+
 # LedgerBERT-Market-Sentiment
 
+This model was introduced in the paper [DLT-Corpus: A Large-Scale Text Collection for the Distributed Ledger Technology Domain](https://huggingface.co/papers/2602.22045). 
+
+The official code repository is available [here](https://github.com/dlt-science/DLT-Corpus).
+
 ## Model Description
 
 ### Model Summary
 
-LedgerBERT-Market-Sentiment is a fine-tuned version of LedgerBERT (https://huggingface.co/ExponentialScience/LedgerBERT) specialized for sentiment analysis of cryptocurrency and DLT-related content. The model classifies text into three market direction sentiment categories: **bullish** (positive market outlook), **bearish** (negative market outlook), and **neutral** (balanced or unclear market direction).
+LedgerBERT-Market-Sentiment is a fine-tuned version of [LedgerBERT](https://huggingface.co/ExponentialScience/LedgerBERT) specialized for sentiment analysis of cryptocurrency and DLT-related content. The model classifies text into three market direction sentiment categories: **bullish** (positive market outlook), **bearish** (negative market outlook), and **neutral** (balanced or unclear market direction).
 
 This model is particularly effective for analyzing cryptocurrency news headlines, social media posts, and other DLT-related content where understanding market sentiment is important.
 
@@ -88,7 +95,7 @@ The dataset provides domain expertise through crowdsourced annotations from cryp
 
 **Note:** News articles are absent from the DLT-Corpus used to pre-train LedgerBERT, making this an out-of-domain generalization test that demonstrates the model's robust language understanding.
 
-For more details on the dataset used for tine-tuning, see: https://huggingface.co/datasets/ExponentialScience/DLT-Sentiment-News
+For more details on the dataset used for fine-tuning, see: https://huggingface.co/datasets/ExponentialScience/DLT-Sentiment-News
 
 ### Training Procedure
 
@@ -161,13 +168,14 @@ for text in texts:
         predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
         predicted_class = predictions.argmax(dim=-1).item()
     
-    # Map to labels (adjust based on your label mapping)
-    labels = ["bearish", "bullish", "neutral"]  # Order may vary
+    # Map to labels based on config.json
+    labels = ["neutral", "bearish", "bullish"]
     sentiment = labels[predicted_class]
     confidence = predictions[0][predicted_class].item()
     
     print(f"Text: {text}")
-    print(f"Sentiment: {sentiment} (confidence: {confidence:.3f})\n")
+    print(f"Sentiment: {sentiment} (confidence: {confidence:.3f})
+")
 ```
 
 ### Batch Processing
@@ -193,7 +201,8 @@ results = classifier(texts, truncation=True, max_length=512)
 
 for text, result in zip(texts, results):
     print(f"Text: {text}")
-    print(f"Sentiment: {result['label']} (score: {result['score']:.3f})\n")
+    print(f"Sentiment: {result['label']} (score: {result['score']:.3f})
+")
 ```
 
 ### Integration with News Feeds
@@ -218,7 +227,8 @@ for entry in feed.entries[:5]:  # Process first 5 entries
     
     print(f"Headline: {title}")
     print(f"Market Sentiment: {result['label']} ({result['score']:.2%})")
-    print(f"Link: {entry.link}\n")
+    print(f"Link: {entry.link}
+")
 ```
 
 ## Citation
@@ -245,7 +255,7 @@ If you use LedgerBERT-Market-Sentiment in your research, please cite:
 
 ### Additional Fine-tuned Models
 
-LedgerBERT can also be fine-tuned for other sentiment dimensions available in the DLT-Sentiment-News dataset (https://huggingface.co/datasets/ExponentialScience/DLT-Sentiment-News):
+LedgerBERT can also be fine-tuned for other sentiment dimensions available in the DLT-Sentiment-News dataset:
 - **Content Characteristics** (liked, disliked, neutral)
 - **Engagement Quality** (important, lol, neutral)