10:50 -11:10 Organizer and workshop introduction, background probe of participants
11:00 – 12:20 Part 1: Talks
11:10 – 11:30 Deep diving academic networks through the lens of OpenAlex (by Noah Mamié)
Abstract: The talk includes introducing OpenAlex as a resource for academic research in various projects. (1) We compare two academic databases (OpenAlex and Microsoft Academic Graph) by benchmarking graph-based algorithms in tasks like node classification. (2) We introduce a graph reasoning framework that is capable of reasoning around the complex topic of deciding on the most worthy individuals to receive the annual Nobel Prize by leveraging graph neural networks.
Slides
11:30 – 11:50 Recommendation system for journals based on ELMo and deep learning (by Mahmoud Hemila)
Abstract: The work evaluates how adequate recommender systems are for the selection of journals that fit to scientific publications. Specifically, several word embedding (word2vec, tf-idf, ELMo) and classification (LR, CNN, RNN, MLP) methods were tested and evaluated against each other in terms of their recommendation accuracy.
Slides
11:50 – 12:10 Large language models and knowledge graph powered academic chatbot (by Susie Xi Rao)
Abstract: We introduce an academic chatbot designed to help identify relevant publications, authors, and affiliations in academia and applied innovation. Leveraging similarity thresholds and query transformations, the chatbot delivers answers that are informed by an academic knowledge base. We address the challenges of efficiency, reproducibility, and interpretability through a combination of rule-based solutions and large language models backed by knowledge-graph embeddings. We have also critically analyzed the output of our chatbot and discussed various future directions of improvement.
Slides
12:10 – 12:20 Preparation for the hands-on session
14:35 –15:35 Part 2: Hands-on
14:30 -15:20 Hands-on session on the academic bot
Slides
15:20 – 15:30 Feedback and exchange