Research Seminar in Semantic Artificial Intelligence
SBWL Kurs V - Knowledge Management
Johann Mitloehner, 2024
Seminar Outline: Simulated Workshop Format
The simulated workshop format is meant to resemble typical situations in academic publishing.
- Topic and abstract submissions to program committee (lecturer)
- Conducting research tasks such as literature overview, data collection, and analysis
- Submission of preliminary paper
- Presentation of paper and results
- Peer review of submissions
- Final paper submission addressing changes proposed by reviewers
Organization
- Students work alone (10 pages) or in groups of two (15 pages).
- Contributions will be formatted as latex documentclass article, so a fictionary proceedings volume can be produced.
- The use of AI tools such as chatGPT for generating the seminar paper is not allowed.
Call for Papers: Workshop on Small-Scale Hybrid symbolic-connectionist AI
Semantic AI combines symbolic and statistical AI
- Symbolic AI -- rules and logical reasoning, e.g. knowledge graphs, logic programming
- Statistical AI -- machine learning to find patterns in the data, e.g. neural nets
Symbolic and statistical/machine learning methods each have their own particular strenghts and weaknesses:
- Symbolic systems provide explicit knowledge encoding and explicit procedures for arriving at results,
but they tend to focus on narrow domains and lack the ability to generalize to other areas.
- Statistical and machine learning systems tend to generalize better but are prone to unreliable
outcomes, particularly when based on large language models.
- The aim of hybrid models is to provide more reliable results based on more explicit information without sacrificing
too much of the power of statistical and machine learning methods.
This simulated workshop aims to bring together various disciplines in Artificial Intelligence,
particularly connectionist Machine Learning, and symbolic approaches based on Semantic Web technologies.
Contributions are welcome on topics dealing with hybrid symbolic-connectionist models with emphasis on small-scale and offline models:
- Sample implementations showing simple approaches to hybrid models
- State of the art Overviews and comparisions of approaches
- Case studies on practical applications, particularly in business domains
Contributions
- P. Maurer, Small Scale AI Finance Management: Combining Budget Rules with Spending Predictions
- Zan Brezovnik and Luka Dovzan Kukic, Hybrid AI System for Team Performance and Lineup Optimization:
Integrating Structured Data and Knowledge Graphs
- Clemens Ploder and Robert Bilato, Visual Question Answering in hybrid symbolic-connectionist systems
Schedule
- 14.10. Intro
→ find relevant overview literature, choose topic: 2 w
- 28.10. Topics assigned
→ find relevant literature for your topic, start work on paper: 5 w
- 02.12. Progress report
→ continue work on paper: 2 w
- 16.12. Prelim. Submission and distribution to reviewers
→ continue work on your paper, review other submissions: 4 w (2 w xmas)
- 13.01. Reviews
→ address reviewer comments in your paper, prepare presentation: 1 w
- 20.01. Presentation
→ prepare final version of paper, including feedback from presentation: 1 w
- 31.01. Final Submission
Related Course Materials
Other Resources