metadata

language:
  - pt
thumbnail: Portugues SBERT for the Legal Domain
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - transformers
datasets:
  - assin
  - assin2
  - stsb_multi_mt
widget:
  - source_sentence: O advogado apresentou as provas ao juíz.
    sentences:
      - O juíz leu as provas.
      - O juíz leu o recurso.
      - O juíz atirou uma pedra.
    example_title: Example 1
model-index:
  - name: SBERTimbau
    results:
      - task:
          name: STS
          type: STS
        metrics:
          - name: PearsonCorrelation
            type: Pearson Correlation
            value: 0.84

rufimelo/Legal-SBERTimbau-sts-large-ma

This is a sentence-transformers model: It maps sentences & paragraphs to a 1024 dimensional dense vector space and can be used for tasks like clustering or semantic search. rufimelo/Legal-SBERTimbau-sts-large-ma-v3 is based on Legal-BERTimbau-large which derives from BERTimbau alrge. It is adapted to the Portuguese legal domain and trained for STS on portuguese datasets.

Usage (Sentence-Transformers)

Using this model becomes easy when you have sentence-transformers installed:

pip install -U sentence-transformers

Then you can use the model like this:

from sentence_transformers import SentenceTransformer
sentences = ["Isto é um exemplo", "Isto é um outro exemplo"]

model = SentenceTransformer('rufimelo/Legal-SBERTimbau-sts-large-ma-v3')
embeddings = model.encode(sentences)
print(embeddings)

Usage (HuggingFace Transformers)

from transformers import AutoTokenizer, AutoModel
import torch


#Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0] #First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('rufimelo/Legal-SBERTimbau-sts-large-ma-v3')
model = AutoModel.from_pretrained('rufimelo/Legal-SBERTimbau-sts-large-ma-v3')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

print("Sentence embeddings:")
print(sentence_embeddings)

Evaluation Results STS

Model	Assin	Assin2	stsb_multi_mt pt
Legal-SBERTimbau-sts-base	0.71457	0.73545
Legal-SBERTimbau-sts-base-ma	0.74874	0.79532	0.82254
Legal-SBERTimbau-sts-base-ma-v2	0.75481	0.80262	0.82178
Legal-SBERTimbau-sts-large	0.76629	0.82357
Legal-SBERTimbau-sts-large-v2	0.76299	0.81121	0.81726
Legal-SBERTimbau-sts-large-ma	0.76195	0.81622	0.82608
Legal-SBERTimbau-sts-large-ma-v2	0.7836	0.8462	0.8261
Legal-SBERTimbau-sts-large-ma-v3	0.7749	0.8470	0.8364
----------------------------------------	----------	----------	----------
BERTimbau base Fine-tuned for STS	0.78455	0.80626	0.82841
BERTimbau large Fine-tuned for STS	0.78193	0.81758	0.83784
----------------------------------------	----------	----------	----------
paraphrase-multilingual-mpnet-base-v2	0.71457	0.79831	0.83999
paraphrase-multilingual-mpnet-base-v2 Fine-tuned with assin(s)	0.77641	0.79831	0.84575

Training

rufimelo/Legal-SBERTimbau-sts-large-ma-v3 is based on Legal-BERTimbau-large which derives from BERTimbau large.

Firstly, due to the lack of portuguese datasets, it was trained using multilingual knowledge distillation. For the Multilingual Knowledge Distillation process, the teacher model was 'sentence-transformers/stsb-roberta-large', the supposed supported language as English and the language to learn was portuguese.

It was trained for Semantic Textual Similarity, being submitted to a fine tuning stage with the assin, assin2 and stsb_multi_mt pt datasets. (batch 8, 5 epochs 'lr': 1e-5)

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False})
)

Citing & Authors

If you use this work, please cite BERTimbau's work:

@inproceedings{souza2020bertimbau,
  author    = {F{\'a}bio Souza and
               Rodrigo Nogueira and
               Roberto Lotufo},
  title     = {{BERT}imbau: pretrained {BERT} models for {B}razilian {P}ortuguese},
  booktitle = {9th Brazilian Conference on Intelligent Systems, {BRACIS}, Rio Grande do Sul, Brazil, October 20-23 (to appear)},
  year      = {2020}
}