数据使用

[howardhsu]

数据使用

[howardhsu]

evaluate的时候用到了官方的code, 需要读取xml格式的文件。

…

On Wed, Jun 13, 2018 at 8:12 AM Lin ***@***.***> wrote: 你好，首先感谢代码开源； 我在阅读代码时，没有发现在训练阶段哪个地方使用了原始的xml格式的数据，好像是.npz的数据，这里跟readme说明是否有出入，谢谢！ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKKt_8eQjK8o52NRvuIltbnX_r34tmV2ks5t8SvdgaJpZM4UmZgC> .

[linWujl]

谢谢，评估的时候用的数据好像是test,也就是测试模型的效果，那在训练阶段用的数据呢，不是官方提供的train的xml数据吗，看了代码好像是读取的npz格式的数据，这个数据是原始训练集处理得来的吗，谢谢。

[howardhsu]

npz是官方xml train文件预处理来的

…

On Thu, Jun 14, 2018 at 8:16 AM Lin ***@***.***> wrote: 谢谢，评估的时候用的数据好像是test,也就是测试模型的效果，那在训练阶段不是用的数据呢，不是官方提供的train的xml数据吗，看了代码好像是读取的npz格式的数据，这个数据是原始训练集处理得来的吗，谢谢。 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKKt_xFXyG8sPiFZ-lJosstbNUw2i5hkks5t8n5NgaJpZM4UmZgC> .

[lsx13]

您好！感谢开源，我用的是python 2.7.12，pytorch 0.3.1，运行的结果F1值跟论文中稍有不同。可否问下，这个开源代码按readme执行下来，是能正好得到论文中描述的结果吗？

[howardhsu]

是的。

…

On Wed, Jul 4, 2018 at 5:41 AM, lsx13 ***@***.***> wrote: 您好！感谢开源，我用的是python 2.7.12，pytorch 0.3.1，运行的结果F1值跟论文中稍有不同。可否问下， 这个开源代码按readme执行下来，是能正好得到论文中描述的结果吗？ — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKKt_x1yYhH1TYBtzLkwtrqqZLoBBKuLks5uDLf6gaJpZM4UmZgC> .

[zhangzhoubin]

你好，非常感谢你的源代码。我运行你的代码时，evaluation.py（laptop）文件 出现java -cp script/eval.jar Main.Aspects不能加载，我查看了一下应该是这个文件没有，还有就是运行evaluation.py（restaurant）出现 IndexError: list index out of range 我试着调试课一下，发现还是没法解决。能帮帮忙吗

[howardhsu]

you forgot to download the official evaluation code/data.

[zhangzhoubin]

data/official_data/pred.xml 这个文件在哪下载，找不见

[howardhsu]

evaluation.py will generate it.

…

On Mon, Sep 17, 2018 at 4:12 AM zhangzhoubin ***@***.***> wrote: data/official_data/pred.xml 这个文件在哪下载，找不见 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKKt_8qKdt4pPzEiSHOARCsbAj-Wjl1Pks5ub2dngaJpZM4UmZgC> .

[linWujl]

Hi, I wonder whether something wrong with the statistic of the dataset semeval-2016 restaurant.
There are 2000 sentences in the file, and 2507 opinion targets with 627 of the opinion targets are 'NULL'. So how to get the number 1743 of aspect terms in your paper.
The test dataset has the same question.
Thanks for your answer.

[howardhsu]

I guess the reason is some terms can belong to multiple category but only need to count once. can you help to check whether you reduplicate those terms?

…

On Wed, Oct 10, 2018 at 10:28 PM Lin ***@***.***> wrote: Hi, I wonder whether something wrong with the statistic of the dataset semeval-2016 restaurant. There are 2000 sentences in the file, and 2507 opinion targets with 627 of the opinion targets are 'NULL'. So how to get the number 1743 of aspect terms in your paper. The test dataset has the same question. Thanks for your answer. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKKt_6HUUteWyqvSd1VyHPv8f1xYHiqjks5ujrrmgaJpZM4UmZgC> .