We kindly invite you to participate in LongEval 2023, a shared task on longitudinal evaluation of NLP models at CLEF 2023.
CALL FOR CONTRIBUTION
LongEval @ CLEF 2023
Longitudinal Evaluation of Models Performance
https://clef-longeval.github.io https://clef-longeval.github.io/
Lab description:
The LongEval aims at identifying the types of models that offer better temporal persistence for NLP tasks on data that evolves across time in both shorter and longer time periods. LongEval is built on a common framework for its Information Retrieval and Text Classification tasks: for one system, we evaluate its efficiency when operating on test data acquired at the same time than the training data, when operating on data acquired shortly after time t (sub-task 1), and when operating on data acquired at time t" ( longer after time t, sub-task 2). For each sub-task of each task, two evaluation measures are proposed: an absolute quality measure, and a relative drop compared to an initial time t test result for each system
LongEval 2023 proposes two tasks:
• Task 1: Information Retrieval. For this task, the data is a sequence of Web document collections and queries, each containing a few million documents and hundreds of queries, provided by Qwant. Relevance assessments are to be computed using a Click Model acquired from real users of the Qwant search engine. As the initial corpus contains only French documents, an automatic translation into English will be provided. • Task 2: Text Classification. For this task, the training data is the TM-Senti sentiment analysis dataset extended with a development set and three human-annotated novel test sets for submission evaluation. TM-Senti is a general large-scale Twitter sentiment dataset in the English language, spanning over a 9-year period from 2013 to 2021. Tweets are labeled for sentiment as either “positive” or “negative”. The annotation is performed using distant supervision based on a manually curated list of emojis and emoticons.
You can register for the task at: https://clef2023-labs-registration.dei.unipd.it/ https://clef2023-labs-registration.dei.unipd.it/
Lab Organizers:
Rabab Alkhalifa, Iman Bilal, Hsuvas Borkakoty, Jose Camacho-Collados, Romain Deveaud, Alaa El-Ebshihy, Luis Espinosa-Anke, Gabriela Gonzalez-Saez, Petra Galuscakova, Lorraine Goeuriot, Elena Kochkina, Maria Liakata, Daniel Loureiro, Harish Tayyar Madabushi, Philippe Mulhem, Florina Piroi, Martin Popel, Christophe Servan, and Arkaitz Zubiaga.
Important dates:
Release of training data: 03/01/2023
Release of test data: 30/04/2023
Runs submissions date: 30/06/2023
LongEval Workshop: during CLEF 2023, Thessaloniki, 18-21 September 2023.