tricktreat / diffusionner

Code for the paper "DiffusionNER: Boundary Diffusion for Named Entity Recognition", accepted at ACL 2023.

👋 Hi there! My name is Yongliang Shen (沈永亮),

• 🔭 I’m currently a fourth-year Ph.D. student advised by Weiming Lu and Yueting Zhuang at Zhejiang University since 2019 (expected to graduate in June 2024). Now I am a research intern at Microsoft Research Asia, mentored by Kaitao Song and Xu Tan.
• 🌱 My research fields include large language models and knowledge graph.
• 📫 How to reach me: syl (at) zju.edu.cn

[Google Scholar] [Semantic Scholar] [Personal Page]

diffusionner's Issues

Empty list for prediction during evaluation

Hello,
Firstly, I would like to thank you for sharing your work.

I've been experimenting with the model but facing issues with inference – I keep getting empty lists for predictions.
I've followed the README and installed dependencies correctly, but the problem persists.
I'm seeking your guidance to understand what could be causing this issue and any specific configurations or test cases that might help.

Thank you in advance!

您能否提供 ACE 数据集预处理代码？

我正在尝试在nested-ner上做一些工作，但是在我处理了ACE数据集后，我发现句子的数量和实体的数量不等于你的论文。你能提供预处理代码吗？
谢谢！

数据集问题

zhmsra_types.json这个文件如何获取？

在网上搜索是需要本地下载，我想问一下这个如果本地下载的话需要怎么该路径呢

关于训练过程的问题

你好我想问一下，在训练过程中，没有类似于unet这样的网络来预测噪声，用这个预测噪声与自己加的高斯噪声作loss吗

一直显示Need 1 GPU for Normal Training, All are busy. Waiting for Free GPU ......，应该怎么解决啊

一直显示Need 1 GPU for Normal Training, All are busy. Waiting for Free GPU ......，应该怎么解决啊

训练结果精确度，F1分数为0，这个会是数据集格式的问题吗？

用自己数据集训练，出现2023-07-19 22:34:43,944 [MainThread ] [INFO ] --- NER on Classification ---
2023-07-19 22:34:43,944 [MainThread ] [INFO ]
2023-07-19 22:34:43,945 [MainThread ] [INFO ] type precision recall f1-score support
2023-07-19 22:34:43,945 [MainThread ] [INFO ] scalename 0.00 0.00 0.00 74.0
2023-07-19 22:34:43,945 [MainThread ] [INFO ]
2023-07-19 22:34:43,945 [MainThread ] [INFO ] micro 0.00 0.00 0.00 74.0
2023-07-19 22:34:43,945 [MainThread ] [INFO ] macro 0.00 0.00 0.00 74.0
2023-07-19 22:34:43,946 [MainThread ] [INFO ] Best Dev-F1 score: 0, achieved at Epoch: 0, Test-F1: None
2023-07-19 22:34:44,286 [MainThread ] [INFO ] Logged in: data/scalener/scalener_train/2023-07-19_22:31:22.027814
2023-07-19 22:34:44,286 [MainThread ] [INFO ] Saved in: data/scalener/scalener_train/2023-07-19_22:31:22.027814

Solution Interpretation

Hii, how do we inference the F1 score which has been provided in the paper for nested NER task and why is the results on the below average side as below when i run as per the steps mentioned in the github for evaluation.

--- NER ---

type precision recall f1-score support

ORG 0.00 0.00 0.00 508
PER 20.42 2.28 4.11 1270
WEA 0.00 0.00 0.00 41
VEH 0.00 0.00 0.00 12
GPE 4.00 0.19 0.35 540
FAC 0.00 0.00 0.00 67
LOC 0.00 0.00 0.00 76

micro 17.96 1.19 2.24 2514
macro 3.49 0.35 0.64 2514

中文ner

支持中文ner吗？

请问有能复现论文中ACE04数据集 F1score88.39的参数吗

请问有能复现论文中ACE2004 F1score88.39的参数吗

How to inferece?

Could you give me an example?Thanks.

how to input text and display the results by DiffusionNER Model?

After the model training is completed, how to input text and display the results? Where do I need to modify the code to implement this functionality?

conll03数据集验证时数据骤降

用conll03数据集训练后的模型做测试的时候F1分数只有70，请教一下是什么问题

3090 24G 显存，出现以下问题，超参已经在作者基础上调小了，怎么解决？

/

which parameter that the threshold is? boundary_threshold ,cls_threshold or entity_threshold ? please tell me the true. Thanks!

In the paper, the razor's threshold=2.5 refers to which parameter's value? The code includes default values such as boundary_threshold = 0.0, cls_threshold = 0.0, entity_threshold = 0.0, but the paper does not specify which parameter actually needs to be modified.

what's the enity recognition generate speed ?

关于训练过程的问题

你好我想问一下，在训练过程中，没有类似于unet这样的网络来预测噪声，用这个预测噪声与自己加的高斯噪声作loss吗

Question of html page

There is a html file in "diffusionner/templates/entity_examples.html"
It should be a flask html template file.
How can i see this web page? I don't know the input variables and cannot print them myself.

OSError: [WinError 123] Incorrect file name, directory name, or volume label syntax。data/ace2004/ace2004_train\\2023-07-29_18:26:47.053675'

How to import diffusionner

Hi, thank u for your work. When I run this code, I notice it has
from diffusionner import input_reader

I did not find this in the code. Is that a python package? or something I misunderstand.