(i) LongEval-Retrieval; (ii) Building a Web-scale privacy-preserving search engine with low budget

Описание: LongEval-Retrieval: French-English Dynamic Test Collection for Continuous Web Search Evaluation
Petra Galuščáková, Université Grenoble Alpes,, France

Building a Web-scale privacy-preserving search engine with low budget: the role of an Information Retrieval PhD
Romain Deveaud, Qwant, France

Talk at the Search Engines course, A.Y. 2022/2023 (https://iiia.dei.unipd.it/education/s...)
MD in Computer Engineering (https://degrees.dei.unipd.it/master-d...)
MD in Data Science (https://datascience.math.unipd.it/)
University of Padua, Italy

----------------------------------------------------------------------
LongEval-Retrieval: French-English Dynamic Test Collection for Continuous Web Search Evaluation

Abstract:
Recent research has shown that as test data becomes more distant in time from the training data, model performance can decrease. This raises important questions for web search engines, such as how they perform as the queries and the collection of documents evolve, and when updates are necessary to keep up with these changes. In my talk, I will introduce the LongEval Lab, which is a new track organized at CLEF 2023. The LongEval Lab will explore how information retrieval copes with evolving document collections over time. I will also describe the LongEval-Retrieval test collection, which has been designed specifically to evaluate information retrieval systems in such dynamic environments. This collection uses data from the privacy-preserving French web search engine Qwant, including real user input and Web documents.

Bio:
Petra Galuščáková is a researcher at the Université Grenoble Alpes, where she works with Lorraine Goeuriot and Philippe Mulhem. She completed her Ph.D. at Charles University in Prague, under the supervision of Pavel Pecina. Her Ph.D. research focused on investigating methods for effective search and navigation in multimedia archives. Following her doctoral studies, Petra worked as a postdoctoral researcher at the University of Maryland, College Park, where she was advised by Doug Oard. Her postdoctoral research was mainly focused on cross-language information retrieval in low resource languages. Her research primarily concentrates on using multiple diverse models for retrieval in large collections under a high degree of uncertainty.


----------------------------------------------------------------------
Building a Web-scale privacy-preserving search engine with low budget: the role of an Information Retrieval PhD

Abstract:
Building and daily running a Web search engine is a challenging project that requires different individuals with a wide range of skills. The international Information Retrieval research community has produced over the last two decades an impressive amount of knowledge related to all components of search engines.
In this talk, I will present how standing on the shoulder of giants can help low budget organisations like Qwant to move forward. I will especially focus on the industry-academy partnership that led to organising the CLEF 2023 LongEval Lab, and the role that Qwant played in building the test collections.

Bio:
Romain Deveaud is a Senior Tech Lead for exploratory and applied research projects at Qwant, a privacy-preserving Web search engine. He is mainly involved in projects that aim at improving ranking & retrieval, query & document understanding, and Web search evaluation. His main goal is to increase the overall ranking quality of Qwant's organic results and improve user satisfaction with the search engine.
Romain holds a PhD in Information Retrieval from Avignon Université, France, and has previously worked with several academic research teams (Glasgow IR group, IRIT, OpenEdition) before joining industry.