https://github.com/cluebenchmark/mobileqa
离线端阅读理解应用 QA for mobile, Android & iPhone
https://github.com/cluebenchmark/mobileqa
albert android bert chinese iphone machine-reading-comprehension nlp qa tensorflow tflite
Last synced: 7 months ago
JSON representation
离线端阅读理解应用 QA for mobile, Android & iPhone
- Host: GitHub
- URL: https://github.com/cluebenchmark/mobileqa
- Owner: CLUEbenchmark
- Created: 2019-11-22T09:34:10.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2022-10-06T03:58:11.000Z (over 3 years ago)
- Last Synced: 2025-05-05T14:51:48.516Z (about 1 year ago)
- Topics: albert, android, bert, chinese, iphone, machine-reading-comprehension, nlp, qa, tensorflow, tflite
- Language: Python
- Homepage:
- Size: 52.6 MB
- Stars: 60
- Watchers: 11
- Forks: 13
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
## MobileQA
[TensorFlow Lite BERT QA Android Example Application](https://github.com/tensorflow/examples/tree/master/lite/examples/bert_qa/android) 与 [tflite-android-transformers](https://github.com/huggingface/tflite-android-transformers) 展示了基于Bert/DistilBERT的离线QA例子，但是只支持英文和安卓设备。
本项目计划实现基于中文的机器阅读理解在手机端的离线应用，并且同时支持安卓和苹果设备。
Targeting to release before Dec 5th. 目标是12月5日前发布。
## 数据集
使用[CMRC 2018 公开数据集](https://github.com/ymcui/cmrc2018/blob/master/README_CN.md)，该数据集是第二届讯飞杯中文机器阅读理解评测所使用的数据。数据集已被计算语言学顶级国际会议[EMNLP 2019](http://emnlp-ijcnlp2019.org/)录用
## 模型
使用 [albert_zh_small](https://github.com/brightmart/albert_zh) 预训练模型，额外加上一层全连接做answer span预测。
* 在CMRC2018数据集的验证集上，max_seq_len为512的模型得分为F1:75.989, EM:52.097, Average:64.038，max_seq_len为384的模型得分为F1:74.781, EM:51.010, Average:62.895
* max_seq_len为512的模型使用tflite转换后大小为18M，经测试，该模型在4线程的安卓手机上推理延时为580ms左右，在单线程条件下为1.4s左右。
* max_seq_len为384的模型使用tflite转换后大小为18M，经测试，该模型在4线程的安卓手机上推理延时为390ms左右，在单线程条件下为930ms左右
* 模型训练与模型转换过程见[bert_cn_finetune-master](https://github.com/CLUEbenchmark/MobileQA/tree/master/bert_cn_finetune-master)
## Android Demo
已完成第一版，详见 [tflite-android-transformers-master](https://github.com/CLUEbenchmark/MobileQA/tree/master/tflite-android-transformers-master)
效果示例：
## IOS Demo
进行中
## Updates
* 2019-11-27: Add first version of Android Demo, details see [tflite-android-transformers-master](https://github.com/CLUEbenchmark/MobileQA/tree/master/tflite-android-transformers-master)
* 2019-11-29: Add first version of MRC model, details see [bert_cn_finetune-master](https://github.com/CLUEbenchmark/MobileQA/tree/master/bert_cn_finetune-master)
## Contribution
如果你感兴趣或希望提供帮助，发送邮件ChineseGLUE@163.com
If you're interested or want to help, send an email to ChineseGLUE@163.com