ERC PhD position in Machine Learning/Natural Language Processing – Miscitations assessment in scientific papers

• Location: LIG laboratory (Grenoble, France)
• Start date: October 2023
• Application deadline: 10 July 2023
• Interview date : 12 July 2023
• Reporting line: Cyril Labbé, Frédérique Bordignon

Keywords: natural language processing, scientific literature, citation accuracy

Context

The NanoBubbles ERC Synergy project is about understanding how, when and why science fails to correct itself. The project focuses on claims made within the field of nanobiology. Project members combine approaches from the natural sciences, computer science, and the social sciences and humanities (Science and Technology Studies) to understand how error correction in science works and what obstacles it faces. For this purpose, we aim to trace claims and corrections through various channels of scientific communication (journals, social media, advertisements, conference programs, etc.) via both qualitative and digital methods.

PhD objectives

In scientific papers, citations are playing a key role. They serve as a mean to acknowledge previous works and position the current work within the context of existing knowldege. When scientific disagreements arise, citations provide readers with clues for understanding and interpreting the work. Most of the time, citations are considered fair in the sense that they accurately reflect the content of the cited paper. Nevertheless, studying the accuracy of citations in various scientific disciplines demonstrate an error rate of 25%-54% (Jergas 2015, Siebers 2000, Kristof 1997, Key 1977). These errors alter the original content and meaning of the cited paper. They can range from minor to major errors in citation accuracy.

Most of the existing works on citation content and citation context analysis focus on determining citation sentiment (Liu 2017, Athar 2011), understanding citation function (Yu 2020, Pride 2019, Bakhti 2018), and identifying critical citation contexts (Te 2022). However, there has been relatively little research on automatically assessing the accuracy of a citation (Palvolic 2020, Agarwal 2023).

The PhD student will have to explore this line of research by first building a dataset, studying and defining a typology of miscitations. When this task is clearly defined, an annotated dataset will need to be built. Then, various natural language processing methods for automatically classifying citations where the context is considered erroneous must be developed.

Skills

• A Master’s degree in Natural Language Processing, computer science, or data science is required.
• Good programming skills in Python, experience with natural language processing tools and frameworks, knowledge of machine learning methods and deep learning technics.
• Ability to communicate and write in English.

Scientific environment

The thesis will be conducted within the Sigma and Getalp teams of the LIG laboratory. The GETALP team has strong expertise and track record in Natural Language Processing, SIGMA team strong expertise in Database and NLP. The recruited person will be welcomed within the teams which offer a stimulating, multinational and pleasant working environment. The means to carry out the PhD will be provided both in terms of missions in France and abroad and in terms of equipment (personal computer, access to the LIG GPU servers). The PhD student will collaborate with several team members involved in the ERC Nanobubbles project, in particular with researchers from the LISIS lab (Marne-la-Vallée, France), University of Sorbonne Paris-Nord as well as researchers from Maastricht University, Radboud Universiteit and University of Twente based in the Netherlands.

Instructions for applying

Applications are expected until July 10th, 2023. They must contain: CV + letter/message of motivation + Master’s academic records + be ready to provide letter(s) of Recommendation; to be addressed to Cyril Labbé (cyril.labbe@imag.fr), Frédérique Bordignon (frederique.bordignon@enpc.fr) and Amira Barhoumi (amira.barhoumi@univ-grenoble-alpes.fr). Applications will be considered case-by-case. It is therefore advisable to apply as soon as possible.

References:

• (Agarwal 2023) Agarwal A, Arafa M, Avidor-Reiss T, Hamoda T, Shah R. Citation Errors in Scientific Research and Publications: Causes, Consequences, and Remedies. The World Journal of Men’s Health, 41.
• (Jergas 2015) Jergas H, Baethge C. Quotation accuracy in medical journal articles-a systematic review and meta-analysis. PeerJ. 2015;3:e1364.
• (Kristof 1997) Kristof C. Accuracy of reference citations in five entomology journals. Am Entomol. 1997;43(4):246-251
• (Key 1977) Key JD, Roland CG. Reference accuracy in articles accepted for publication in the Archives of Physical Medicine and Rehabilitation. Arch Phys Med Rehabil. 1977;58(3):136-137.
• (Siebers 2000) Siebers R, Holt S. Accuracy of references in five leading medical journals. Lancet. 2000;356(9239):1445
• (Te 2022) Te S, Barhoumi A, Lentschat M, Bordignon F, Labbé C, Portet F. Citation Context Classification: Critical vs Non-critical. In proceedings of the Third Workshop on Scholarly Document Processing. 2022:49-53.
• (Liu 2017) Liu H. Sentiment analysis of citations using word2vec. 2017;CoRR, abs/1704.00177.
• (Athar 2011) Athar A. Sentiment analysis of citations using sentence structure-based features. In Proceedings of the ACL 2011 Student Session. 2011:81–87.
• (Bakhti 2018) Bakhti K, Niu Z, Yousif A, Nyamawe A. Citation Function Classification Based on Ontologies and Convolutional Neural Networks. 2018:105–115.
• (Palvolic 2020) Pavlovic V, Weissgerber T, Stanisavljevic D, Pekmezovic T, Garovic V, Milic N and CITE Investigators. How accurate are citations of frequently cited papers in biomedical literature?. bioRxiv preprint doi: https://doi.org/10.1101/2020.12.10.419424
• (Pride 2019) Pride D, Knoth P, Jozef Harag J. Act: An annotation platform for citation typing at scale. In 2019 ACM/IEEE Joint Conference on Digital Libraries (JCDL). 2019:329–330.
• (Yu 2020) Yu W, Yu M, Zhao T, Jiang M. Identifying referential intention with heterogeneous contexts. 2020:962–972.