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

这是一个 sentence-transformers 模型，它能将句子和段落映射到一个 768 维的密集向量空间，可用于聚类或语义搜索等任务。rufimelo/rufimelo/Legal-BERTimbau-sts-base-ma 基于 Legal-BERTimbau-base，而后者源自 BERTimbau 大模型。该模型适用于葡萄牙语法律领域，并在葡萄牙语数据集上针对语义文本相似度（STS）进行了训练。

🚀 快速开始

安装依赖

使用此模型，你需要安装 sentence-transformers：

pip install -U sentence-transformers

运行示例代码

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：能够将句子和段落映射到 768 维的密集向量空间。
• 多任务适用：可用于聚类、语义搜索等任务。
• 领域适配：针对葡萄牙语法律领域进行了适配。
• 多数据集训练：在多个葡萄牙语数据集上进行了训练。

📦 安装指南

若要使用该模型，需安装 sentence-transformers：

pip install -U sentence-transformers

💻 使用示例

基础用法（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)

高级用法（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)

📚 详细文档

评估结果 STS

训练过程

rufimelo/Legal-BERTimbau-sts-base-ma-v2 基于 Legal-BERTimbau-base，而后者源自 BERTimbau 基础模型。

首先，由于葡萄牙语数据集的缺乏，该模型采用多语言知识蒸馏进行训练。在多语言知识蒸馏过程中，教师模型为 'sentence-transformers/paraphrase-xlm-r-multilingual-v1'，假定支持的语言为英语，要学习的语言为葡萄牙语。

该模型针对语义文本相似度进行了训练，并在 assin、assin2 和 stsb_multi_mt pt 数据集上进行了微调。

完整模型架构

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})
)

引用与作者

如果你使用了此工作，请引用以下文献：

@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}
}

🔧 技术细节

数据集

训练方法

• 多语言知识蒸馏：由于葡萄牙语数据集的缺乏，使用多语言知识蒸馏进行训练，教师模型为 'sentence-transformers/paraphrase-xlm-r-multilingual-v1'。
• 微调：在多个葡萄牙语数据集上进行微调，以适应语义文本相似度任务。

模型架构

模型基于 SentenceTransformer 构建，包含 Transformer 层和 Pooling 层，具体架构如下：

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})
)

评估指标

使用 Pearson 相关性作为评估指标，在多个数据集上进行评估，以衡量模型在语义文本相似度任务上的性能。

代码实现

在代码实现中，使用了 Sentence-Transformers 和 HuggingFace Transformers 库，通过简单的代码即可加载和使用模型，具体示例如下：

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)