Current Schedule: Link
| Date | Presenter | Paper | Why should we care | Appear in | Repo | Blogs |
|---|---|---|---|---|---|---|
| 29/03/21 | Dima Galat | Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation | Good introduction to student-teacher models and forcing the vector mapping | EMNLP 2020 | 1 2 | 1 |
| 12/04/21 | Duy Khuu | In defense of pseudo-labeling: An uncertainty-aware pseudo-label selection framework for semi-supervised learning | Introduction to pseudo-labeling and its potential issues along with a proposal for a better performing variant | ICLR 2021 | ||
| 19/04/21 | Andrew Law | ConStance: Modeling Annotation Contexts to Improve Stance Classification | Introduction to issues with manual data annotation and proposes a model for reasoning about annotations with different contexts | EMNLP 2017 | ||
| 26/04/21 | Quyu Kong | PhD conclusion talk dry-run | ||||
| 03/05/21 | Marian-Andrei Rizoiu | Adversarial vulnerabilities of human decision-making | Adversarial learning to guide sequential human decision making | PNAS'20 | 1 | |
| 10/05/21 | Rohit Ram | The Social Dynamics of Language Change in Online Networks | Tracks language change through social influence and determines importance of (diverse) features to social influence using Parametric Hawkes | SocInfo'16 | ||
| 17/05/21 | Frankie Yuan | All You Need is “Love”: Evading Hate Speech Detection | Simple but effective adversarial attacks on word and character based hate speech classification models. paper link | AISec 2018 | ||
| 24/05/21 | Thomas Willingham | Generating Datasets with Pretrained Language Models | leveraging large well-trained generative language models (e.g GPT2) for creating labelled data to fine-tune semantic embedders and comes with code and all | Computing Research Repository | 1 | |
| 31/05/21 | – | No group reading. Public holiday in Canberra (Reconciliation Day) | ||||
| 07/06/21 | Pio Calderon | Learning Multivariate Hawkes Processes at Scale | Scalable and exact approach to train multivariate Hawkes processes in large-scale domains | WWW21 | 1 | |
| 14/06/21 | – | No group reading. Public holiday (Queen's Birthday) | ||||
| 21/06/21 | Quyu Kong | Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks | Meta-learning allows to adapt to new tasks quickly with a few samples | ICML'17 | 1 | |
| 28/06/21 | Rohit Ram, Andrew Law, and Marian-Andrei Rizoiu | 3MT Series | ||||
| 05/07/21 | Andrew Law | Characterizing Online Discussion Using Coarse Discourse Sequences | Characterizes different types of online discussions (Q&A, arguments etc.) through analyzing patterns of discourse and produces a classifier for predicting discourse acts. | ICWSM 2017 | ||
| 12/07/21 | Rohit Ram | A General Method to Find Highly Coordinating Communities in Social Media through Inferred Interaction Links | ||||
| 19/07/21 | Duy Khuu | The Benchmark Lottery | Overview of current issues related to ranking models and benchmarks, along with proposals for how we can produce better benchmarks | Submitted to NeurIPS 2021 | ||
| 26/07/21 | Duy Khuu, Pio Calderon, and Quyu Kong | 3MT Series | ||||
| 02/08/21 | Pio Calderon | Is Gang Violent Crime More Contagious than Non‑Gang Violent Crime? | Modelling the spatiotemporal dynamics of gang/non-gang crime contagion with multivariate Hawkes | Journal of Quantitative Criminology | ||
| 09/08/21 | Frankie Yuan | Detecting Troll Behavior via Inverse Reinforcement Learning: A Case Study of Russian Trolls in the 2016 US Election | Detection and analysis of Russian troll behaviour and insentives using inverse reinforcement learning. | ICWSM 2020 | 1 | |
| 16/08/21 | Dima Galat | All That's ‘Human’ Is Not Gold: Evaluating Human Evaluation of Generated Text | An actionable proposal on how evaluations should be conducted | ACL21 | ||
| 23/08/21 | Thomas Willinghanm, Dima Galat, and Frankie Yuan | 3MT Series | ||||
| 30/08/21 | Marian-Andrei Rizoiu | TextAttack: A Framework for Adversarial Attacks, Data Augmentation, and Adversarial Training in NLP | A Python package to perform adversarial attacks for a number of tasks. It also allow adversarial training. | EMNLP’20 Demo | 1 | |
| 06/09/21 | Thomas Willingham | Disrupting Model Training with Adversarial Shortcuts | An approach to deliberately poisoning datasets for privacy or competitive purposes | ICML2021 | ||
| 13/09/21 | Andrew Law | Honours mid-term talk dry run | ||||
| 20/09/21 | – | No group reading. UTS Recharge day leave | ||||
| 27/09/21 | Quyu Kong | Identifying Coordinated Accounts on Social Media through Hidden Influence and Group Behaviours | Detect coordinated accounts by jointly modeling account activities and hidden group behaviours based on Temporal Point Processes (TPP) and Gaussian Mixture Model (GMM). | KDD'21 | 1 | |
| 04/10/21 | – | No group reading. Public holiday (Labour Day) | ||||
| 11/10/21 | Callum Pastuszak | Using a Personality-Profiling Algorithm to Investigate Political Microtargeting: Assessing the Persuasion Effects of Personality-Tailored Ads on Social Media | Demonstrates how personality profiling algorithms can be used to target political ads at individuals on an online setting increasing their receptiveness to vote | Sage Journal | ||
| 18/10/21 | Duy Khuu | Online discussion threads as conversation pools: predicting the growth of discussion threads on reddit | Generative approach to model conversation pools on Reddit | Computational and Mathematical Organization Theory | 1 | |
| 25/10/21 | Thomas Willinghanm | Honours final talk dry run | ||||
| 01/11/21 | Duy Khuu | Honours final talk dry run | ||||
| 08/11/21 | Andrew Law | Honours final talk dry run | ||||
| 15/11/21 | Pio Calderon | Simple discrete-time self-exciting models can describe complex dynamic processes: A case study of COVID-19 | A discrete-time Hawkes process to model mortality counts during two phases of the COVID19 outbreak | PLOS One | ||
| 22/11/21 | – | No reading group! Andrei decided to conveniently skip his turn and go on leave. | ||||
| 29/11/21 | Rohit Ram | CA1 Dry Run | ||||
| 06/12/21 | Quyu Kong | The Surprising Performance of Simple Baselines for Misinformation Detection | Extensive evaluations on common benchmark datasets for misinformation detection. | WWW'21 | 1 |