ZJB's BERT-Tiny Multilingual Sentiment Model (v2.0)

模型概述 / Model Overview

这是一个基于 BERT-Tiny 架构微调的双语（中文/英文）情感分析模型。该模型使用约5000条高质量的中英双语数据进行训练，能够以极高的准确率和置信度判断文本的情感倾向（正面/负面）。

This is a bilingual (Chinese/English) sentiment analysis model fine-tuned on the BERT-Tiny architecture. Trained on approximately 5000 high-quality bilingual data pairs, it can judge the emotional tendency (positive/negative) of text with high accuracy and confidence.

模型详情 / Model Details

• 模型类型 (Model Type): BERT-Tiny (约4M参数 / ~4M Parameters)
• 微调者 (Fine-tuned by): ZJB
• 基础模型 (Base Model): prajjwal1/bert-tiny
• 训练数据 (Training Data): ~5000 条自定义中英双语句子 / ~5000 custom bilingual sentences.
• 最大序列长度 (Max Sequence Length): 64 tokens
• 输入格式 (Input Format): 原始文本 / Raw text。
• 输出格式 (Output Format): 标签 (LABEL_1 为正面, LABEL_0 为负面) 及置信度 / Label (LABEL_1 for positive, LABEL_0 for negative) with confidence score.

如何使用 / How to Use

您可以直接使用 transformers 库的 pipeline API： You can use the pipeline API from the transformers library directly:

from transformers import pipeline

classifier = pipeline("text-classification", model="zjb522/bert-tiny-zjb-sentiment-v2")
result = classifier("The atmosphere and service of this restaurant are top-notch!")
print(result)
# [{'label': 'LABEL_1', 'score': 0.9958407282829285}]