An explanation classifier which classifies input-output pairs of Citeomatic (neural based citation recommendation system) into a relevant citation or not. Used BERT for embedding the abstracts of input and output papers. This model added explanations to “blackbox” decision process in neural models, and outperformed the original CBR (case-based reasoning) model both in terms of accuracy and macro F1 score.
More …
This is an assignment from B552 “Knowledge Based AI.” In this assignment, I implemented a rule based system to derive animal taxonomy from given information. The rules are in the form (name, left hand side (LHS), right hand side (RHS)), and the classifier performs depth first search to match the LHS of rules and apply substitution on the RHS to derive new knowledge. The classifier ends its execution when no more new knowledge is derived by the algorithm.
More …
I modeled Chinese abbreviation process as a tagging problem and achieved F1-score of 0.83 on testing.
More …
14 Jan 2021
Thanks for visiting my site. I’m Kazuki, a master student at Indiana University studying Computational Linguistics.
Please go to About for my casual introduction.
Please go to Backgrounds for my backgrounds.
Please go to Projects for my projects.
Click here for the code.
More …