Sources for our paper titled "Predicting Helpful Posts in Open-Ended Discussion Forums: A Neural Architecture", NAACL 2019.

If you use these sources, please consider citing the paper:

@inproceedings{halder-etal-2019-predicting,
    title = "Predicting Helpful Posts in Open-Ended Discussion Forums: A Neural Architecture",
    author = "Halder, Kishaloy and Kan, Min-Yen and Sugiyama, Kazunari",
    booktitle = "Proceedings of the 2019 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)",
    month = jun,
    year = "2019",
    pages = "3148--3157",

• README.md: This file.
• relevance_novelty.py: Main file to run the code.
• utilities.py: Contains some utility functions.
• constant.py: Contains values for some parameters.

1. It requires pre-trained glove embedding file. It can be downloaded from here. https://nlp.stanford.edu/projects/glove/ Once you have extracted the .zip file, use the full path of the embedding file (named glove.6B.100d.txt) in constants.py as the value of the WORD_EMBEDDING_FILE variable.

2. You also need to provide path of the data file as the value of POSTS_FILE in constants.py. This file should be a tab (\t) separated file having the following data columns:

• postID: A unique ID for each post.
• postText: Textual content of the target post to be classified.
• OPost: Textual content of the original post in the same thread as that of the target post.
• helpfulCount: A binary value (0 or 1) depicting the two classes (non-helpful and helpful respectively).
• context_[0-19]: 20 separate columns each with a contextual post from the same thread as that of the target post. Note that the 19th (context_19) column would always be the penultimate post in the same thread. As for example, if the target post is the 5th in the thread, then context_14 would be the first post in the thread, context-15 the second, similarly context_19 would be the fourth (and thus penultimate) in the thread. Other columns (context_[0-13]) would be left empty.

Once you have these in place, you can train the model using the following:

python3 relevance_novelty.py

All the outputs will be logged in a file in the ./logs folder.

You would need the following installed in your system to run the sources.

• Python 3.x
• keras
• pandas
• sklearn
• tqdm