jianliu-ml / eventcausalityidentification

1. Put EventStoryLine, CausalTM, EventCausalityData, SemEval, and FrameNet datasets in the current directory.

2. Run read_document.py to build all intra-sentence examples for EventStoryLine.

3. Run preprecess_*.py to generate BERT-based examples for difference datasets and settings.

4. Run train_mix.py to train the model, with different settings.

eventcausalityidentification's Issues

Which part of the is explictly using the conceptnet?

这个版本的代码里似乎没有找到显式利用常识图谱进行知识检索的部分？我没仔细看数据准备部分的代码，是包含在那里面了吗？

Code for Table 1

Can the author release the code for Table 1?
In your paper, the last two topics are used as the development set, and cross-validation is used to train the model. However, I have not seen any details for this part.

In addition, I have implemented the BERT model myself, but the results are far different from what you reported. Can the author also provide the code for BERT baseline?

Could you please add more comments to show us how to reproduce the experiment results?

I found it really difficult to actually run the code.
Which experiment is the train_topic() method in train_mix.py mean for?
I guess it is the Cross-Topic Adaptation (5.1, Table 4).
But I am interested in the experiments in Table 1. Where is the method for that experiment?
Also all the other experiments in the paper.

The topic_c() method in util.py does not work

I found that topic_c() method cannot load the data.pickle, where the data structure does not match.
Can you check it?
Can anyone run the code?

Hello, I would like to ask if data_knowledge.pickle file does not have application

请问你是做的二分类还是三分类？

我看你的代码里面，处理数据和计算指标的时候都是当成二分类来做的