pwnslinger / cyberbullying-nlp Goto Github PK

@RoyaSalek for our work, SVM with hinge function to max the unlabeled data margin would be sufficient
we can borrow the implementation from here:
https://github.com/oroszgy/semisup-learn/blob/master/semisup_learn/methods/qns3vm.py

http://pages.cs.wisc.edu/~jerryzhu/pub/sslicml07.pdf

Implementing the XLNet model for text classification sentiment analysis.

https://arxiv.org/abs/1906.08237

• marginal data labeling analysis
  • PCA / SVD to see if we can get a reliable results
• marginal data random sampling
• accuracy
• Bias-Variance Decomposition
  • cross-validations
  • seperation of train/teset data

https://web.stanford.edu/class/archive/cs/cs224n/cs224n.1194/reports/custom/15787612.pdf

• integrating more word embeddings into the code
  • Elmo
  • ULM-Fit
• BERT

@sogol-golafshan
http://arno.uvt.nl/show.cgi?fid=148294

Please read this one and make you understand what's going on.

• removing preprocessing from the baseline code
• relacing acronyms with the collected list of mappings
• voter classifier
  • @sogol-golafshan clean code, run it

@mahdi-whip needs to implement a voter classifier which is easy. send me a PR on GitHub when you are done with it.

link: https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.RANSACRegressor.html

We asked @parisahjb to implement:

• ROC curve (feasible)
• learning rate
• AUC curve
• Confusion Matrix
• deliver an excel spread sheet which contains the following:
  • please come up with a parser to put the data in CSV format

• @sogol-golafshan delivered you the dataset. (issue: #4)
• annotation results in xls