bangliu / storyforest

This repository implements the system described in "Growing Story Forest Online from Massive Breaking News"

Story Forest

This reporsitory implements Story Forest system for event clustering and story generation.

Please cite the following paper for usage:

@inproceedings{liu2017growing,
  title={Growing Story Forest Online from Massive Breaking News},
  author={Liu, Bang and Niu, Di and Lai, Kunfeng and Kong, Linglong and Xu, Yu},
  booktitle={Proceedings of the 2017 ACM on Conference on Information and Knowledge Management},
  pages={777--785},
  year={2017},
  organization={ACM}
}

For event clustering, see "eventlayer/EventLayer"; for story generation, see "story layer/StoryLayer". The main functions in the two files show how to run the whole process.

Notice: currently, we haven't release the dataset due to copyright issues. But we will release a sample dataset asap.

storyforest's Issues

test_data

Hi, Thank you for your beautiful work. I have a question: Where is the test_data?

Library upgrade edu.ualberta.storyteller

After installing every library with maven, I still encounter that certain functions from the edu.ualberta.storyteller library cannot be found. I do not know if this is a mistake on my side but it seems to be something on the library of edu.ualberta.storyteller.

Exception in thread "main" java.io.FileNotFoundException: model\svm_model.tencent-news

Are only unigram keywords considered in the event clustering algorithm ?

We have setup the code provided in this repo. We were experimenting with both unigram keywords and bigram keywords but we were not able to observe any bigram keywords in the output of the event clustering pipeline. Does this indicate that bigram keywords are not considered in event clustering algorithm?

Whether English dataset can be used in the model? 这个模型能不能分享一下（svm_model.20ng）

can you release a sample dataset?

Is this all the code

I did not find many parts of the code, such as word segmentation and keyword extraction

Can the values for parameters (such as segTitle, topic, Timestamp) be provided for English Corpora ?

There are parameters (such as segTitle, topic, Timestamp) missing in the data loading step for English corpora in the code. Has the Story Forest pipeline not been run for the English Corpora ?