This is the official repository of "IR From Bag-of-words to BERT and Beyond through Practical Experiments", a CIKM 2021 full-day tutorial with PyTerrier and OpenNIR search toolkits.
This is an updated edition of our previous ECIR 2021 Tutorial
- Part 1: Classical IR: indexing, retrieval and evaluation
- Part 2: Modern Retrieval Architectures: PyTerrier data model and operators, towards re-rankers and learning-to-rank
- Part 3: Contemporary Retrieval Architectures: Neural re-rankers such as BERT, EPIC, ColBERT
- Part 4: Recent Advances beyond the classical inverted index: neural inverted index augmentation, nearest neighbor search, dense retrieval
Our tutorial will run twice on Friday 5th November 2021, with live lab sessions.
Suggested viewing:
- Western Europe: Run 1
- North/South America: Run 2
- Asia and Oceana: Run 1 (parts 1 & 2), followed by Run 2 (parts 3 & 4)
Run 1: Aligned with Western Europe, partially accessible to Americas or Asia/Oceana
| London (GMT) | Content | Live | Beijing (GMT+x) | Brisbane (GMT+x) |
|---|---|---|---|---|
| 0900-1000 | part 1 slides | ✔️ | ||
| 1000-1030 | part 1 lab | ✔️ | ||
| 1030-1100 | break | ☕ | ||
| 1100-1200 | part 2 slides | ✔️ | ||
| 1200-1230 | part 2 lab | ✔️ | ||
| 1230-1330 | break | ☕ | ||
| 1330-1430 | part 3 slides | ✔️ | ||
| 1430-1500 | part 3 lab | ✔️ | ||
| 1500-1530 | break | ☕ | ||
| 1530-1630 | part 4 slides | ✔️ | ||
| 1630-1700 | part 4 lab | ✔️ |
Run 2: Aligned with North/South America, partially accessible to Asia/Oceana
| SF (GMT-x) | Content | Live | Beijing (GMT+x) | Brisbane (GMT+x) |
|---|---|---|---|---|
| 1000-1100 | part 1 slides | 🎥 | ||
| 1100-1130 | part 1 lab | ✔️ | ||
| 1130-1200 | break | ☕ | ||
| 1200-1300 | part 2 slides | 🎥 | ||
| 1300-1330 | part 2 lab | ✔️ | ||
| 1330-1430 | break | ☕ | ||
| 1430-1530 | part 3 slides | 🎥 | ||
| 1530-1600 | part 3 lab | ✔️ | ||
| 1600-1630 | break | ☕ | ||
| 1630-1730 | part 4 slides | 🎥 | ||
| 1730-1800 | part 4 lab | ✔️ |
We are thankful to Luca and Eugene for supporting additional live labs, and for CIKM 2021 volunteers, for being able to support this schedule.
- Sean MacAvaney, University of Glasgow, UK
- Craig Macdonald, University of Glasgow, UK
- Nicola Tonellotto, University of Pisa, IT
Contributors to Live Labs:
- Luca Soldini, Amazon
- Eugene Yang, Johns Hopkins University
- PyTerrier: [Github] [Documentation]
- OpenNIR: [Github] [Documentation]
- PyTerrier_ColBERT: [Github]
- PyTerrier_T5: [Github]
- PyTerrier_doc2query: [Github]
- PyTerrier_DeepCT: [Github]
- PyTerrier_ANCE: [Github]
- PyTerrier_DeepImpact [Github]
If you make using of any of these slides, notebooks, or additional PyTerrier plugins, please cite our tutorial abstract:
@inproceedings{ecir2021-tut-bow2b,
author = {MacAvaney, Sean and Macdonald, Craig and Tonellotto, Nicola},
title = {IR From Bag-of-words to BERT and Beyond through Practical Experiments: A CIKM 2021 tutorial with PyTerrier and OpenNIR},
booktitle = {Proceedings of CIKM 2021},
year = {2021}
}