CLEARumor: ConvoLving ELMo Against Rumors @ RumorEval 2019

This repository contains the code used in the following paper:

Ipek Baris, Lukas Schmelzeisen, and Steffen Staab (2019). “CLEARumor at SemEval-2019 Task 7: ConvoLving ELMo Against Rumors”. In: SemEval@NAACL-HLT. Association for Computational Linguistics, pp. 1105–1109. [ACL Anthology] [arXiv] [Poster]

If you use this in your work, please cite as:

@inproceedings{DBLP:conf/semeval/BarisSS19,
  author    = {Ipek Baris and
               Lukas Schmelzeisen and
               Steffen Staab},
  title     = {{CLEAR}umor at {S}em{E}val-2019 Task 7: {C}onvo{L}ving {ELM}o
               {A}gainst {R}umors},
  booktitle = {SemEval@NAACL-HLT},
  pages     = {1105--1109},
  publisher = {Association for Computational Linguistics},
  year      = {2019},
  url       = {https://www.aclweb.org/anthology/S19-2193},
}

Dependencies

Python 3.6+ is required.

• See the PyTorch Getting Started page for how to install it.
• Install AllenNLP:

  pip3 install allennlp

• Install the tokenizer for Twitter and Reddit posts:

  pip3 install git+https://github.com/erikavaris/tokenizer.git

Data

• Place the ELMo weights and options files elmo_2x4096_512_2048cnn_2xhighway_5.5B_weights.hdf5 and elmo_2x4096_512_2048cnn_2xhighway_5.5B_options.json in the data/external subdirectory. They can be obtained from the ELMo website.
• Place the training, test data, and evaluation files rumoureval-2019-training-data.zip, rumoureval-2019-test-data.zip, and home_scorer_macro.py in the data/external/ subdirectory. They can be obtained from the RumorEval 2019 competetion page.
• If you also want to evaluate on the final evaluation data from RumorEval, place the final-eval-key.json file in the data/external subdirectory. So far, this file has only been shared privately with all participants in the competition. It is under an embargo for a couple of months, but will be released publicly later.

Contacts

If you have any questions regarding the code or the employed machine learning architectures, please, don't hesitate to contact the authors or report an issue.

• Lukas Schmelzeisen, [email protected], @lschmelzeisen
• Ipek Baris, [email protected], @ipekbrs
• Institute for Web Science and Technoloogies, University of Koblenz-Landau

License

All code is license under the Apache License Version 2.0.