Norbert Fuhr's recommendations for gaining scientific knowledge from experiments:
1. Do not use MRR or MAP;
2. Instead of relative improvements, regard the effect size!
3. For multiple significance tests, use a correction, such as Bonferoni or Tukey's HSD (NB comparing only to the 2nd best method does not help!)
4. There are no significant improvements for re-usable test collections! (hypotheses have to be formulated before the work)
#SIGIR2020
More at: https://sigir.org/sigir2020/
SIGIR 2020 highlights for people in European time zones:
Wednesday 29 July
8:30--9:30h CEST -- Keynote VI by Dacheng Tao: How Deep Learning Works for Information Retrieval
9:40-11:40h CEST -- Full paper session 8B, Multi-modal Retrieval and Ranking
13:00-14:00h CEST -- Playback Keynote V by Elizabeth Churchill: From Information to Assistance
15:00-18:00h CEST -- Business Meeting & Closing
SIGIR 2020 highlights for people in European time zones:
Tuesday 28 July
8:30--9:30h CEST -- Keynote IV by Norbert Fuhr: Proof by Experimentation? Towards Better IR Research
9:40-11:40h CEST -- Full paper session 5C, Information Access and Filtering
12:45-13:45h CEST -- Playback Keynote III by Ellen Voorhees: Coopetition in IR Research
14:00-16:00h CEST -- Posters/Demos/Social event
16:00-18:00h CEST -- Full paper session 6C Context-aware Modeling
SIGIR 2020 highlights for people in European time zones:
Monday 27 July
8:30--9:30h CEST -- Keynote II by Zong-Ben Xu: On Presuppositions of Machine Learning: A Meta Theory
9:40-12:00h CEST -- Full paper session 2B, User Behavior and Experience
13:00-14:30h CEST -- Women in IR Panel
14:45-15:45h CEST -- Playback Keynote I by Geoffrey Hinton: The Next Generation of Neural Networks
16:00-18:00h CEST -- Full paper session 3B, Learning to Rank
Information Retrieval and Its Sister Disciplines. (arXiv:1912.02346v1 [cs.IR]) http://arxiv.org/abs/1912.02346
I ditched Google for DuckDuckGo. Here's why you should too:
https://www.wired.co.uk/article/duckduckgo-google-alternative-search-privacy
Beyond word embeddings: learning entity and concept representations from large scale knowledge bases http://link.springer.com/10.1007/s10791-018-9340-3
#IR researcher in NL or B (or keen on visiting Amsterdam):
Better register soon: https://www.dir2019.nl may run out of capacity, so if you are not quick...
Universal Transformers: a transformer approach by Mostafa Dehghani and colleagues that also uses a recurrent feedback loop. Very interesting. #SIKS
Slides for all lectures at the #Siks #IR course:
https://www.informagus.nl/events/siks-ir-2019.html
Here, you also find the information for the group experiment of this afternoon.
Slides of @Claudia Hauff's SIKS course presentation "Machine Learning for #IR" available online!
https://docs.google.com/presentation/d/1Jiu5ZkRJVrrFmwtGJg9elClDah-f1eKXCEkQXORrchk/edit?usp=sharing
Full house for the #Siks #IR Course for PhD students in Utrecht, NL.
Faegheh Hasibi @fhasibi presenting the lecture on Entity Oriented Search now.
Morning program included Claudia Hauff on Learning to Rank and other Machine learning for #IR, and Hinda Haned on FAT* related topics such as algorithmic bias.
"The merits of Universal Language Model Fine-tuning for Small Datasets - a case with Dutch book reviews"
https://arxiv.org/abs/1910.00896
Work by Benjamin van der Burgh @LIACS, evaluating the effectiveness of ULMFiT for small training sets
Vacancies!
15 fully-funded PhD Positions on Domain-Specific Search (EU Marie Curie Action project)
http://ifs.tuwien.ac.at/dossier/phd_positions.html
I am hiring for project 6 and 7, addressing transparency and explainability in legal search.
We're hiring a PhD Candidate for Transfer Learning for Federated Search!
https://www.ru.nl/english/working-at/vacature/details-vacature/?recid=1064653