Knowledge Representation and Reasoning Seminar

Lehrveranstaltung mit SWS 0/2/0 (Vorlesung/Übung/Praktikum) in WS 2019

Dozent

Steffen Hölldobler
Emmanuelle Dietz

Umfang (SWS)

0/2/0

Module

Leistungskontrolle

Mündliche Prüfung
Referat

Information

The next seminar will take place on Thursday, 16.1.2020 at 14:50 in APB2026.

Knowledge Representation and Reasoning Seminar

The seminar will be about the most recent results on the Winograd Schema Challenge.

The requirements for the KRR Seminar are as follows:

You need to be at least a minimum of five students that want to participate
You attend all talks during the semester
You select one of the papers presented below and communicate your choice to Emmanuelle Dietz until 14.11.2019
You give a presentation of 30 minutes about the chosen paper in January 2020
You send (a preliminary version of) your presentation slides until 16.12.2019 to Emmanuelle Dietz

Schedule

The seminar meetings will take place on thursdays, 5.DS (14:50 - 16:20, starting on 24.10.2019) in room APB2026.

17.10.19 initial meeting
24.10.19 Presentation of the topics
29.10.19 Graph matching, theory and SAT implementation by Orianne Laura Bargain (this talk will take place on Tuesday, 10:30)
07.11.19 SCF2 - an Argumentation Semantics for Rational Human Judgments on Argument Acceptability by Marcos Cramer
14.11.19 Abduction in a neuro-symbolic system by Andrzej Gajda
21.11.19 How to make a presentation in LaTeX. Template slides in Beamer for presentations are online. You can find them here how to give a talk I how to give a talk II
28.11.19 TE-ETH: Lower Bounds for QBFs of Bounded Treewidth by Markus Hecher, joint work with Johannes Fichte and Andreas Pfandler (this talk will take place at 13:00 together with the KBS seminar in APB3027, see also the recent GI newsletter for their guest commentary in German)
5.12.19 Human Syllogistic Reasoning: Towards Predicting Individuals' Reasoning Behavior based on Cognitive Principles by Robert Schambach (joint work with Emmanuelle Dietz)
12.12.19 Justifying All Differences Using Pseudo-Boolean Reasoning by Marcos Cramer
17.12.19 Graph matching, theory and SAT implementation by Stephan Gocht (this talk will take place on Tuesday, 15:00 in APB2028)
19.12.19 Feedback on handed in presentations
09.01.20 Google’s T5 - A Unified Text-to-Text Transformer by Patrick Wienhöft
16.01.20 Machine Learning approaches towards WSC by Vidya Chandrashekar (part I) and by Abhiram Uppoor (part II)
23.01.20 Human Baseline for Commonsense Reasoning Tasks by Vishwanath Hugar and COPA: Choice of Plausible Alternatives by Aldo Kurmeta
30.01.20 How Reasonable are Common-Sense Reasoning Tasks by Lukas Gerlach

-->

Commonsense Reasoning

This seminar will be about commonsense reasoning in AI, and the Winograd Schema Challenge, an alternative to the Turing Test.

Topics

What is commonsense reasoning?
- sources
  - wikipedia
  - http://commonsensereasoning.org/
  - Davis; Marcus (2015). "Commonsense reasoning". Communications of the ACM. Vol. 58 no. 9. pp. 92–103.
  - McCarthy, J. (1959). "Programs with Common Sense". Proceedings of the Teddington Conference on the Mechanization of Thought Processes (pp. 75--91), London: Her Majesty's Stationary Office.

Winograd Schema Challenge (WSC)
- sources
  - wikipedia
  - http://commonsensereasoning.org/
  - Levesque, Davis, and Morgenstern (2012). "The Winograd Schema Challenge". KR
  - Levesque (2013). "On Our Best Behaviour". IJCAI Research Excellence Award Presentation
  - Morgenstern, Davis, and Ortiz (2016). "Planning, Executing, and Evaluating the Winograd Schema Challenge". AI Magazine

Human Baseline for Commonsense Reasoning Tasks
- sources
  - Davis, Morgenstern, Oriz (2016). Human tests of materials for the Winograd SchemaChallenge 2016
  - Bender, Establishing a Human Baseline for the Winograd Schema Challenge. MAICS 2015
  - Nangia and Bowman, A Conservative Human Baseline Estimate for GLUE: People Still (Mostly) Beat Machines
  - Nangia, Bowma, Human vs. Muppet: A Conservative Estimate of Human Performance on the GLUE Benchmark, 2019

Machine Learning approaches towards WSC
- sources
  - Trichelair et al. (2018). On the Evaluation of Common-Sense Reasoning in Natural Language Understanding
  - Trinh and Le (2018). A Simple Method for Commonsense Reasoning
  - Radford et al. (2019). Language Models are Unsupervised Multitask Learners
  - Ruan, Zhu, Ling, Liu, Wei .Exploring Unsupervised Pretraining and Sentence Structure Modeling for Winograd Schema Challenge
  - Kocijan, Cretu, Camburu, Yordanov, Lukasiewicz (2019). A Surprisingly Robust Trick for Winograd Schema Challenge

A critical view on Commonsense Reasoning Tasks
- possible sources
  - Trichelair, Emami, Trischler, Suleman, Cheung. "How Reasonable are Common-Sense Reasoning Tasks: A Case-Study on the Winograd Schema Challenge and SWAG"