This repository contains a pre-trained Chinese in-car natural language understanding (NLU) model based on BERT. It is designed for automotive voice assistant scenarios and supports joint intent classification and slot filling for task-oriented dialogue.

The model files include:

config.json: model configuration
model.safetensors: model weights
tokenizer.json, tokenizer_config.json, vocab.txt, special_tokens_map.json: tokenizer assets
onnx/model.onnx: optional ONNX export for optimized inference
onnx/label_map.json: intent and slot label definitions

Model Description

A BERT-based NLU model for in-vehicle dialogue understanding. It is tailored for automotive instruction interpretation and can predict both user intent and relevant slot entities from Chinese commands.

Key capabilities:

intent classification for driving-related user requests
slot extraction for entities such as destinations, controls, music, weather, phone contacts, and climate settings
optimized runtime formats including safetensors and ONNX

模型简介

该模型基于 BERT 架构构建，面向车载语音助手场景。它支持中文指令的意图识别与槽位抽取，用于理解用户在驾驶过程中的自然语言请求。

功能特点：

支持多类别意图预测
支持槽位实体抽取
提供 safetensors 权重及 ONNX 推理格式

Intended Use

This model is intended for research and development of Chinese automotive NLU systems, including:

in-car virtual assistants
navigation and media control
climate and window management
telephony and weather queries

The repository is structured as a model package rather than a complete application. Use your own inference or evaluation scripts to load the model and tokenizer.

使用说明

该目录包含用于推理的模型权重与分词器文件，可直接加载到 Hugging Face Transformers 风格的推理代码中。

Basic usage example:

python

from transformers import BertTokenizerFast, BertForTokenClassification
import torch

model_dir = "."
tokenizer = BertTokenizerFast.from_pretrained(model_dir)
model = BertForTokenClassification.from_pretrained(model_dir)

inputs = tokenizer("今天天气怎样？", return_tensors="pt")
outputs = model(**inputs)

如果使用 ONNX 推理，可加载 onnx/model.onnx 进行加速部署。

文件结构

config.json
model.safetensors
tokenizer.json
tokenizer_config.json
vocab.txt
special_tokens_map.json
onnx/model.onnx
onnx/label_map.json

训练与评估

本仓库并未包含完整训练脚本。本 README 主要描述模型卡信息与部署说明。如需训练或微调，请参考您自己的 BERT NLU 训练流程。

免责声明

本模型适用于车载中文自然语言理解任务。若将其用于生产环境，请先进行充分评估与测试。

mhhyoucom/bert-model-car-assistant

bert-model-car-assistant

Model Overview