NSP-CitiLink is a fine-tuned BERT model specialized in Text Segmentation for municipal council meeting minutes. The model uses Next Sentence Prediction (NSP) to identify topic boundaries in long-form documents, making it particularly effective for segmenting administrative and governmental meeting minutes.

Try out the model: Hugging Face Space Demo

Key Features

🎯 Specialized for Meeting Minutes: Fine-tuned on Portuguese municipal council meeting minutes
⚡ Fast Inference: Efficient BERT-base architecture for real-time segmentation
📊 High Accuracy: Achieves BED F-measure score of 0.79 on CitiLink dataset
🔄 Sentence-Level Segmentation: Identifies topic boundaries at sentence granularity

Model Details

Base Model: neuralmind/bert-base-portuguese-cased
Architecture: BERT with Next Sentence Prediction head
Parameters: 110M
Max Sequence Length: 512 tokens
Fine-tuning Dataset: CitiLink (Portuguese Municipal Meeting Minutes)
Fine-tuning Method: Focal Loss with boundary-aware weighting
Training Framework: PyTorch + Transformers

How It Works

The model predicts whether two consecutive sentences belong to the same topic (label 0: "is_next") or represent a topic transition (label 1: "not_next"). By applying this classifier sequentially across all sentence pairs in a document, it identifies topic boundaries.

python

Sentence A: "Pelo Senhor Presidente foi presente a reunião a ata n.º 28 de 20.12.2023."
Sentence B: "Ponderado e analisado o assunto o Executivo Municipal deliberou por unanimidade aprovar a ata n.º 28 de 20.12.2023."
→ Prediction: Same Topic (confidence: 76%)

Sentence A: "Ponderado e analisado o assunto o Executivo Municipal deliberou por unanimidade aprovar a ata n.º 28 de 20.12.2023."
Sentence B: "Não houve processos e requerimentos diversos a apresentar."
→ Prediction: Topic Boundary (confidence: 82%)

Usage

Quick Start with Transformers

python

from transformers import AutoTokenizer, AutoModelForNextSentencePrediction
import torch

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("liaad/Citilink-NSP-Segmentation-pt")
model = AutoModelForNextSentencePrediction.from_pretrained("liaad/Citilink-NSP-Segmentation-pt")

# Prepare input
sentence_a = "Pelo Senhor Presidente foi presente a reunião a ata n.º 28 de 20.12.2023."
sentence_b = "Ponderado e analisado o assunto o Executivo Municipal deliberou por unanimidade aprovar a ata n.º 28 de 20.12.2023."


# Tokenize
inputs = tokenizer(sentence_a, sentence_b, return_tensors="pt")

# Predict
with torch.no_grad():
    outputs = model(**inputs)
    logits = outputs.logits
    probs = torch.softmax(logits, dim=1)
    
# Interpret results
is_next_prob = probs[0][0].item()
not_next_prob = probs[0][1].item()

print(f"Is Next (same topic): {is_next_prob:.3f}")
print(f"Not Next (topic boundary): {not_next_prob:.3f}")

if not_next_prob > 0.5:
    print("🔴 Topic boundary detected!")
else:
    print("🟢 Same topic continues")

Limitations

Domain Specificity: Best performance on administrative/governmental meeting minutes
Language: Optimized for Portuguese; English performance may vary
Document Length: Designed for documents with 10-50 segments
Context Window: Limited to 512 tokens per sentence pair
Ambiguous Boundaries: May struggle with subtle topic transitions

Model Card Contact

For questions or feedback, please open an issue in the model repository.

License

This project uses a custom dual-license based on AGPL v3.

See the full license terms here: LICENSE

inesctec/CitiLink-NSP-Segmentation-pt

NSP-CitiLink: Linear Text Segmentation for Municipal Meeting Minutes

Model Description