Abusive Language Detection with Graph Convolutional Networks
- Submitting institution
-
University of Cambridge
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 10354
- Type
- E - Conference contribution
- DOI
-
10.18653/v1/N19-1221
- Title of conference / published proceedings
- Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies
- First page
- 2145
- Volume
- 1
- Issue
- -
- ISSN
- -
- Open access status
- -
- Month of publication
- June
- Year of publication
- 2019
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
3
- Research group(s)
-
-
- Citation count
- -
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- The paper demonstrates that framing abusive language classification in Tweets as a semi-supervised graph convolution problem, where the graph both represents authors and their Tweets, improves on a baseline using logistic regression. The new formulation allows the resulting model to exploit homophily in social networks as well as the observation that specific authors are responsible for many abusive Tweets without the need for explicit author profiling in the training dataset, yielding the best performance on this dataset at the time of publication.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -