trait-relations-and-co-occurences

Code and full results in csv files for experiments in: Assessing the Limits of the Distributional Hypothesis in Semantic Spaces: Trait-based Relational Knowledge and the Impact of Co-occurrences

Code is split into directories:

• classifier-code (for training and evaluating with SVM classifiers)
• datasets-code (code used to develop trait-concept datasets)
• embbeding-code (code used for traing w2v models)
• removal-code (code used for extracting and replacing co-occurrence instances)
• wiki-code (code use to pre-process Wikipedia data)

Results directory includes:

• results.csv (results from main experiment with multilabel SVM classifiers)
• results-binary.csv (results from binary SVM classifiers)

The code, in all its disgraceful glory, is shared for transparency rather than for other people to use. Such a fate no soul deserves.

@inproceedings{and22-dist-hyp,
    title = "Assessing the Limits of the Distributional Hypothesis in Semantic Spaces: {T}rait-based Relational Knowledge and the Impact of Co-occurrences",
    author = "Anderson, Mark and Camacho Collados, Jose",
    booktitle = "To appear in proceedings of *SEM 2022: The Eleventh Joint Conference on Lexical and Computational Semantics",
    month = jul,
    year = "2022",
    address = "Seattle",
    publisher = "Association for Computational Linguistics",
}