Legal BERTimbau Sts Base Ma V2

🚀 rufimelo/Legal-BERTimbau-sts-base-ma

這是一個 sentence-transformers 模型，它能將句子和段落映射到一個 768 維的密集向量空間，可用於聚類或語義搜索等任務。rufimelo/rufimelo/Legal-BERTimbau-sts-base-ma 基於 Legal-BERTimbau-base，而後者源自 BERTimbau 大模型。該模型適用於葡萄牙語法律領域，並在葡萄牙語數據集上針對語義文本相似度（STS）進行了訓練。

🚀 快速開始

安裝依賴

使用此模型，你需要安裝 sentence-transformers：

pip install -U sentence-transformers

運行示例代碼

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

model = SentenceTransformer('rufimelo/Legal-BERTimbau-sts-base-ma-v2')
embeddings = model.encode(sentences)
print(embeddings)

✨ 主要特性

• 向量映射：能夠將句子和段落映射到 768 維的密集向量空間。
• 多任務適用：可用於聚類、語義搜索等任務。
• 領域適配：針對葡萄牙語法律領域進行了適配。
• 多數據集訓練：在多個葡萄牙語數據集上進行了訓練。

📦 安裝指南

若要使用該模型，需安裝 sentence-transformers：

pip install -U sentence-transformers

💻 使用示例

基礎用法（Sentence-Transformers）

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

model = SentenceTransformer('rufimelo/Legal-BERTimbau-sts-base-ma-v2')
embeddings = model.encode(sentences)
print(embeddings)

高級用法（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-BERTimbau-sts-base-ma-v2')
model = AutoModel.from_pretrained('rufimelo/Legal-BERTimbau-sts-base-ma-v2')

# 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)

📚 詳細文檔

評估結果 STS

訓練過程

rufimelo/Legal-BERTimbau-sts-base-ma-v2 基於 Legal-BERTimbau-base，而後者源自 BERTimbau 基礎模型。

首先，由於葡萄牙語數據集的缺乏，該模型採用多語言知識蒸餾進行訓練。在多語言知識蒸餾過程中，教師模型為 'sentence-transformers/paraphrase-xlm-r-multilingual-v1'，假定支持的語言為英語，要學習的語言為葡萄牙語。

該模型針對語義文本相似度進行了訓練，並在 assin、assin2 和 stsb_multi_mt pt 數據集上進行了微調。

完整模型架構

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})
)

引用與作者

如果你使用了此工作，請引用以下文獻：

@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}
}

@inproceedings{fonseca2016assin,
  title={ASSIN: Avaliacao de similaridade semantica e inferencia textual},
  author={Fonseca, E and Santos, L and Criscuolo, Marcelo and Aluisio, S},
  booktitle={Computational Processing of the Portuguese Language-12th International Conference, Tomar, Portugal},
  pages={13--15},
  year={2016}
}

@inproceedings{real2020assin,
  title={The assin 2 shared task: a quick overview},
  author={Real, Livy and Fonseca, Erick and Oliveira, Hugo Goncalo},
  booktitle={International Conference on Computational Processing of the Portuguese Language},
  pages={406--412},
  year={2020},
  organization={Springer}
}
@InProceedings{huggingface:dataset:stsb_multi_mt,
title = {Machine translated multilingual STS benchmark dataset.},
author={Philip May},
year={2021},
url={https://github.com/PhilipMay/stsb-multi-mt}
}

🔧 技術細節

數據集

訓練方法

• 多語言知識蒸餾：由於葡萄牙語數據集的缺乏，使用多語言知識蒸餾進行訓練，教師模型為 'sentence-transformers/paraphrase-xlm-r-multilingual-v1'。
• 微調：在多個葡萄牙語數據集上進行微調，以適應語義文本相似度任務。

模型架構

模型基於 SentenceTransformer 構建，包含 Transformer 層和 Pooling 層，具體架構如下：

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})
)

評估指標

使用 Pearson 相關性作為評估指標，在多個數據集上進行評估，以衡量模型在語義文本相似度任務上的性能。

代碼實現

在代碼實現中，使用了 Sentence-Transformers 和 HuggingFace Transformers 庫，通過簡單的代碼即可加載和使用模型，具體示例如下：

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

model = SentenceTransformer('rufimelo/Legal-BERTimbau-sts-base-ma-v2')
embeddings = model.encode(sentences)
print(embeddings)

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-BERTimbau-sts-base-ma-v2')
model = AutoModel.from_pretrained('rufimelo/Legal-BERTimbau-sts-base-ma-v2')

# 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)