ECIR 2023 Dublin – conference report

Last week I attended the 45th edition of the European Conference on Information Retrieval (ECIR) which was hosted in Dublin, Ireland. Given that this was only my second in-person conference I have ever attended it was great to bump into several familiar faces throughout the week. One reason for that could be that this year’s ECIR was the biggest ECIR of all time with a record-breaking number of more than 400 registered attendees. The high level of interest was evident on-site, as more than 300 people attended the keynote talks and receptions.

On Sunday, the first conference day started with seven different workshops in parallel and all attendees had the chance to choose one of the following: Text2Story, GeoExT, Legal IR, ROMCIR, BIAS, BIR and ALTARS. I decided for the ROMCIR (Reducing Online Misinformation through Credible Information Retrieval) workshop as its topics are closely linked to my own area of work and some of my colleagues presented their research there. For me it was great to learn more about things like the reduction of harm on social media and false health related information on the web.

In the evening of day one the welcome reception took place in the Teeling Distillery with plenty of Whiskey and opportunities to talk to people. This event was a first taste of the welcoming Irish culture that you could experience throughout the whole five conference days.

The second day (Monday) of the conference was also the first day of the main event. The sessions of the main program were almost all in non-parallel which allowed everybody to experience the conference the same way. One of my highlights was Mounia Lalmas’ keynote on personalization at Spotify. In her talk she shared some interesting insights on how recommendation and search are handled at Spotify. Apart from that the day was packed with many interesting paper presentations as well as poster and demo sessions in the late afternoon.

program on the third day followed a similar structure to the previous day. However, the keynote talk by William Yang Wang from UC Santa Barbara on the challenges and opportunities of large language models (LLMs) for question answering stood out as the most remarkable part for me. William Yang Wang is also the recipient of this year’s Karen Spärck Jones award. Moreover, this conference day was also when I presented my poster. I found it inspiring to have so many insightful discussions and receive feedback on my work. The poster sessions were well-attended, and it seemed like other presenters had a similar experience to mine. The highlight of this conference day was undoubtedly the Banquet in the evening. The event featured excellent food, company, and engaging talks, making it a memorable experience.

The final day of the main conference was packed with a diverse set of talks, although there were no more poster or demo sessions. I primarily focused on attending talks related to the CLEF shared tasks, and I must say that I appreciate the organizers’ effort to creating opportunities for tackling interesting though challenging problems in collaboration with the community. The day concluded with a closing session that drew a full house of attendees. During this session, Glasgow was announced as the host city for ECIR 2024, and the best paper awards were presented:

Best Student Short Paper: Investigating Conversational Search Behavior for Domain Exploration by Phillip Schneider et al.

Best Short Paper: Doc2Query–: When Less is More by Mitko Gospodinov et al.

Best Student Long Paper: Improving Video Retrieval Using Multilingual Knowledge Transfer by Avinash Madasu et al.

Best Long Paper: Temporal Natural Language Inference: Evidence-Based Evaluation of Temporal Text Validity by Taishi Hosokawa et al.

On Thursday, there was the industry day along with several tutorials, with about 60 attendees. I participated in the industry day talks and particularly remember the presentation by Jakub Zavrel’s (Zeta Alpha) on the combination of neural search and language model prompting, as well as Simon Smith’s presentation (Ripjar Ltd) on preventing financial crime through machine-curated profiles.

Overall, this year’s ECIR conference featured a variety of interesting themes, and my main takeaway is the upcoming trend of combining generative LLMs and classical information retrieval. ChatGPT was also prominently featured throughout the conference. Other topics that caught my attention include the various use cases for graph data structures, such as in recommendation or network community detection, as well as discussions surrounding fairness in information retrieval. These are just a few examples that stood out to me.

ECIR: Long days of work require long evenings to recover

It’s also worth noting that the organizers did an excellent job in preparing the social events and selecting the venue. The social events provided a great opportunity to connect with people and experience the Irish culture. The venue, the Radisson Blue Royal Hotel in Dublin, was located close to the city center making it easy to explore the city or reach authentic Irish pubs in short time. In addition, there was a great lunch buffet in the hotel’s restaurant every day and the premises made it possible to host smaller sessions in parallel (workshops) and the main sessions without overlap ensuring that everyone could attend their preferred sessions.



About Gregor Donabauer
Gregor Donabauer

Gregor Donabauer is a PhD student and research assistant at the chair of Information Science at the University of Regensburg (Germany). Before that he completed his Bachelor degree in Information Science/Business Information Systems and his Master degree in Information Science, both at the University of Regensburg. His main interests are on Natural Language Processing and Machine Learning.

Leave a Reply

You must be logged in to post a comment.