Reading Group

A reference document used to keep track of the reading group meetings. Pull requests are welcome. Meeting Link

Schedule 2021

DatePresenterPaperWhy should we careAppear inRepoBlogs
29/03/21Dima GalatMaking Monolingual Sentence Embeddings Multilingual using Knowledge DistillationGood introduction to student-teacher models and forcing the vector mappingEMNLP 20201 21
12/04/21Duy KhuuIn defense of pseudo-labeling: An uncertainty-aware pseudo-label selection framework for semi-supervised learningIntroduction to pseudo-labeling and its potential issues along with a proposal for a better performing variantICLR 2021
19/04/21Andrew LawConStance: Modeling Annotation Contexts to Improve Stance ClassificationIntroduction to issues with manual data annotation and proposes a model for reasoning about annotations with different contextsEMNLP 2017
26/04/21Quyu KongPhD conclusion talk dry-run
03/05/21Marian-Andrei RizoiuAdversarial vulnerabilities of human decision-makingAdversarial learning to guide sequential human decision makingPNAS'201
10/05/21Rohit RamThe Social Dynamics of Language Change in Online NetworksTracks language change through social influence and determines importance of (diverse) features to social influence using Parametric HawkesSocInfo'16
17/05/21Frankie YuanAll You Need is “Love”: Evading Hate Speech DetectionSimple but effective adversarial attacks on word and character based hate speech classification models. paper linkAISec 2018
24/05/21Thomas WillinghamGenerating Datasets with Pretrained Language Modelsleveraging large well-trained generative language models (e.g GPT2) for creating labelled data to fine-tune semantic embedders and comes with code and allComputing Research Repository1
31/05/21No group reading. Public holiday in Canberra (Reconciliation Day)
07/06/21Pio CalderonLearning Multivariate Hawkes Processes at ScaleScalable and exact approach to train multivariate Hawkes processes in large-scale domainsWWW211
14/06/21No group reading. Public holiday (Queen's Birthday)
21/06/21Quyu KongModel-Agnostic Meta-Learning for Fast Adaptation of Deep NetworksMeta-learning allows to adapt to new tasks quickly with a few samplesICML'171
28/06/21Rohit Ram, Andrew Law, and Marian-Andrei Rizoiu3MT Series
05/07/21Andrew LawCharacterizing Online Discussion Using Coarse Discourse SequencesCharacterizes different types of online discussions (Q&A, arguments etc.) through analyzing patterns of discourse and produces a classifier for predicting discourse acts.ICWSM 2017
12/07/21Rohit RamA General Method to Find Highly Coordinating Communities in Social Media through Inferred Interaction Links
19/07/21Duy KhuuThe Benchmark LotteryOverview of current issues related to ranking models and benchmarks, along with proposals for how we can produce better benchmarksSubmitted to NeurIPS 2021
26/07/21Duy Khuu, Pio Calderon, and Quyu Kong3MT Series
02/08/21Pio CalderonIs Gang Violent Crime More Contagious than Non‑Gang Violent Crime?Modelling the spatiotemporal dynamics of gang/non-gang crime contagion with multivariate HawkesJournal of Quantitative Criminology
09/08/21Frankie YuanDetecting Troll Behavior via Inverse Reinforcement Learning: A Case Study of Russian Trolls in the 2016 US ElectionDetection and analysis of Russian troll behaviour and insentives using inverse reinforcement learning.ICWSM 20201
16/08/21Dima GalatAll That's ‘Human’ Is Not Gold: Evaluating Human Evaluation of Generated TextAn actionable proposal on how evaluations should be conductedACL21
23/08/21Thomas Willinghanm, Dima Galat, and Frankie Yuan3MT Series
30/08/21Marian-Andrei RizoiuTextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLPA Python package to perform adversarial attacks for a number of tasks. It also allow adversarial training.EMNLP’20 Demo1
06/09/21Thomas WillinghamDisrupting Model Training with Adversarial ShortcutsAn approach to deliberately poisoning datasets for privacy or competitive purposesICML2021
13/09/21Andrew LawHonours mid-term talk dry run
20/09/21No group reading. UTS Recharge day leave
27/09/21Quyu KongIdentifying Coordinated Accounts on Social Media through Hidden Influence and Group BehavioursDetect coordinated accounts by jointly modeling account activities and hidden group behaviours based on Temporal Point Processes (TPP) and Gaussian Mixture Model (GMM).KDD'211
04/10/21Callum Pastuszak
11/10/21No group reading. Public holiday (Labour Day)
18/10/21Duy Khuu
25/10/21Pio Calderon
01/11/21Frankie Yuan
08/11/21Dima Galat
15/11/21Rohit Ram
22/11/21Marian-Andrei Rizoiu
29/11/21Thomas Willinghanm
06/12/21Duy Khuu
13/12/21Pio Calderon
20/12/21Marian-Andrei Rizoiu
27/12/21Quyu Kong
03/01/22Frankie Yuan
10/01/22Andrew Law
17/01/22Rohit Ram
24/01/2022Dima Galat