Academic Intent Detection based on Fine-tunned Bert Model

**Overview** This listing describes the the fine-tuned model for intent detection based on the BERT model. The dataset used for fine-tuning was generated with academic content from a set of subjects and the AWS Bedrock Claude 3 Haiku. The results was then evaluated by LLM judges. During fine-tuning, the model achieved an accuracy of 95.5% on the 20-80% dataset. The model is currently registred and served as an agent in Databricks. **Use cases** - Academic Tutor **Product details** Language: Portuguese **Additional Insights** Example of intent detection: - user_message: "Bom dia" -> intent: "saudacao"

搜集汇总

背景与挑战

背景概述

该数据集基于BERT模型微调，用于学术意图检测，使用葡萄牙语学术内容生成，在评估中达到95.5%的准确率。模型已注册并部署在Databricks平台，适用于学术辅导等场景。