This special issue of International Journal on Digital Libraries (IJDL) brings together a selection of high-quality papers that were presented at the 24th International Conference on Asia–Pacific Digital Libraries (ICADL 2022). ICADL is an annual international interdisciplinary conference that, together with JCDL and TPDL conference, is one of the three top venues for connecting digital library, computer science and library as well as information science communities. The 24th International Conference on Asia–Pacific Digital Libraries was held in Ha Noi, Vietnam from November 30 to December 2, 2022. ICADL 2022 was planned as a forum for researchers to exchange ideas and discuss together across domains for an innovative digital information environment. The theme of the conference was: “From Born-Physical to Born-Virtual: Augmenting Intelligence in Digital Libraries”. As best papers, the articles in this special issue of IJDL have gone through a rigorous review process. The top highly evaluated and scored papers at ICADL 2022 were first selected and then invited to be published at the special issue after their extension. They were then peer-reviewed by experts to help the authors make substantive changes for improving their work. The four papers contained in this special issue that emerged from our editorial process are as follows:

The first paper titled “Stance Prediction with a Relevance Attribute to Political Issues in Comparing the Opinions of Citizens and City Councilors” and authored by Ko Senoo, Yohei Seki, Wakako Kashino, Atsushi Keyaki and Noriko Kando approaches the task of comparing stances and their of citizens and city councilors on political issues. The authors have annotated tweets from citizens as well as city council minutes based on four labels to construct a dataset for fine-tuning LLMs. Their approach was based on multitask learning and using T5 as a prediction model. The authors could detect some interesting opinion gaps such as differences in evaluation of vaccination regimes.

The second paper titled “Exploiting the Untapped Functional Potential of Memento Aggregators Beyond Aggregation” and authored by Mat Kelly discusses the concept of Memento aggregation in web archives identifying the state-of-the-art in pure server-side aggregators (Time Travel) and user deployable aggregators (MemGator). It emphasizes the importance of treating web requests agnostically and outlines foundational assumptions of an aggregator, exploring expectations regarding the treatment of clients’ requests and the statelessness of aggregators. Further to this it includes the description of modifications to two open-source tools, MemGator and Mink, as proof-of-concepts to enhance Memento aggregation features.

The third paper titled “User vs institutional perspectives of metadata and searching: an investigation of online access to cultural heritage content during the COVID-19 pandemic”, and authored by Ryan Colin Gibson, Sudatta Chowhury, and Gobinda Chowdury investigates the mismatch between the way content was indexed by the curators and searched for by users based on digital content hosted by two large national cultural heritage institutions – National Museums of Scotland (NMS) and National Galleries of Scotland (NGS). Based on the log data analysis, which consists of access patterns to both NMS and NGS websites, and also combined with user study, the authors conclude that user- and context-specific guidelines could be beneficial in ensuring the aspects considered most important by consumers are indexed, thereby producing more relevant search results. As a result, this should lead to an improvement on the individual’s experience when searching for information.

The last paper, “Human-in-the-loop Latent Space Learning for Bibrecord-based Literature Management”, and authored by Shingo Watanabe, Hiroyoshi Ito, Masaki Matsubara, Atsuyuki Morishima presents an interactive tool for personal literature management based on bibliographic records. The proposed method represents the relationship between bibliographic records as a heterogeneous graph. Then, it applies a human-in-the-loop learning iteration based on a graph convolutional encoder-decorer model for learning the latent space in the user’s mind. Ten researchers from humanities, science, and engineering domains participated in experiments and tested the management system, which showed the effectiveness of the proposed method in placing unknown documents at the user’s desired position.

The four articles that have gone through the editorial process for this special issue of IJDL represent a broad spectrum of high-quality work in our communities. We hope that the readers will find these papers as interesting and insightful as we did.