zliucr / mixed-language-training Goto Github PK

View Code? Open in Web Editor NEW

32.0 5.0 6.0 14.85 MB

Attention-Informed Mixed-Language Training for Zero-shot Cross-lingual Task-oriented Dialogue Systems (AAAI-2020)

License: MIT License

Python 100.00%

cross-lingual zero-shot task-oriented-dialogue low-resource multilingual nlu dst

mixed-language-training's People

Stargazers

Watchers

Forkers

currywu123 gentaiscool meryemmhamdi1 xixirupan aaron2bin arthurminovsky

mixed-language-training's Issues

关于vocab和embed

你好，可以对nlu_emb的各个文件做下说明吗，比如使用的什么样的embedding，感谢~

你好，这个代码什么时候能出来

DST在加载数据时，把验证集也用来训练了？

def load_data(params, dialogue_ontology, mapping=None):
    pri_turns_train = load_woz_data("data/dst/dst_data/tok_woz_train_en.json", "en", dialogue_ontology, mapping=mapping)
    pri_turns_val = load_woz_data("data/dst/dst_data/tok_woz_validate_en.json", "en", dialogue_ontology, mapping=mapping)
    val_count = len(pri_turns_val)
    pri_turns_train = pri_turns_train + pri_turns_val[0:int(0.75 * val_count)]

    tgt_pri_turns_val = load_woz_data("data/dst/dst_data/tok_woz_validate_"+params.trans_lang+".json", params.trans_lang, dialogue_ontology, mapping=mapping)
    tgt_pri_turns_test = load_woz_data("data/dst/dst_data/tok_woz_test_"+params.trans_lang+".json", params.trans_lang, dialogue_ontology, mapping=mapping)
    
    return pri_turns_train, tgt_pri_turns_val, tgt_pri_turns_test

看起来这里把英语的验证集的75%也加到训练集了。这样做是可以的么？如果这样做可以的话，那么为什么这个比例是75%？
可不可以把英语的测试集也加到训练集里？

The code version dependency

Hi Zihan,

Thank you for sharing the code of the paper. I'm trying to reproduce it now. I install torch==0.4.0 with your settings, but I have some problems. The error is as follows:

And I try the latest version torch==1.10.0, but another error is as follows:

Do you have any experience in solving them?

Regards,
Guojun

Asking for the source code of Transformer/Multi. BERT/XLM

Hi Zihan,

Thanks for contributing your source code of the paper https://arxiv.org/pdf/1911.09273.pdf.

In this source code, I have not find the Transformer/Multi. BERT/XLM models, as they are the state-of-the-art models as reported in the paper.

Could you please share the models or let us know how to reproduce the reported model.

Regards,
Jiahuan

zliucr / mixed-language-training Goto Github PK

mixed-language-training's People

Stargazers

Watchers

Forkers

mixed-language-training's Issues

关于vocab和embed

你好，这个代码什么时候能出来

DST在加载数据时，把验证集也用来训练了？

The code version dependency

Asking for the source code of Transformer/Multi. BERT/XLM

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent