Conference Agenda (All times are shown in Eastern Daylight Time)

This study explored whether the application of generative search (GS) techniques—including RAG, COT, and RAT—is superior to traditional keyword search (KS) techniques in digital museum scenarios. We constructed a prototype search system using 800 painting collections from the digital collection repository platform of the Palace Museum in Beijing. Sixteen users performed factual, explanatory, and exploratory search tasks while evaluating five search methods, including KS and various GS optimizations. The experiment results reveal that GS performs better than KS, despite challenges with user satisfaction and search accuracy because of vague responses and occasional hallucinations. Among GS techniques, RAT—integrating chain-of-thought reasoning with document retrieval—achieved the best overall performance. The findings help overcome current search challenges in digital cultural heritage and provide practical guidance for advancing the intelligent development of online museums.

This paper describes an exploratory study applying an AI chatbot to the transformation of archival collections into linked data representations in a community archive setting. Through qualitative analysis of research diaries, we analyze the experiences of novice data curators using ChatGPT to develop item-level metadata records conformant to linked data standards based on a collection of informally digitized items. Findings indicate that users find ChatGPT useful for a few, discrete archival processing tasks with limited types of items, but ChatGPT’s significant inconsistencies impede systematized workflows. Our findings shed light on anticipated benefits and challenges of using generative AI to create linked data from digital collections in resource-limited settings, relevant to archival institutions and community organizations seeking low-barrier opportunities to make their collections more accessible and interconnected online.

Cultural heritage resource metadata is the foundation and precious asset for GLAM institutions to provide knowledge services which enables users to efficiently search relevant collection information. However, current GLAM institutions (e.g. museums), face significant challenges to gather comprehensive high-quality collection metadata. Motivated by the complementary advantages of multi-modal large language models (MLLMs) and pretrained small ones (MPSMs), we proposed “AGGM”, a model fusion approach for automatic metadata enrichment including two key components: MPSM-based module for demonstration detection and MLLM-based module for prediction calibration. The merit of AGGM is to fully leverage the powerful semantic understanding capabilities to generate accurate results based on MLLMs in limited computation cost, and also exert the strength of MPSM domain-specific knowledge to obtain informative demonstrations. The experimental results showed that AGGM outperformed baseline models in two regular metadata generation tasks, demonstrating enormous potential of this proposed model fusion approach in automatic generation of GLAM metadata.

We share results from a mixed-method, three-year grant-funded investigation into whether Library, Archive, and Museum (LAM) organizations can integrate crowdsourced transcription data into content management systems (CMSs) or databases, and whether doing so makes primary sources such as diaries and letters discoverable and accessible for blind people or those with low vision (BLV). Our research questions are concerned with the ease and ability of LAMs to integrate crowdsourced data, LAM and academic crowdsourcing project owners’ beliefs and attitudes about such data, and the discoverability and accessibility of crowdsourced data for BLV users. This paper shares findings from a survey and two interview types conducted with 12 LAM Partner organizations to understand practitioner attitudes about crowdsourced data quality, and data ingest tools, systems, and processes. We applied Evaluation and Magnitude Coding to our LAM Partner data. We hypothesized that crowdsourced data might be over-collected and underutilized due to complex data structures, distrust in volunteer crowdsourcing, and the absence of usable technical infrastructure and metadata standards to support data integration into LAM CMSs. We found significant heterogeneity in data ingest processes but higher rates of successful ingest than expected: 10 out of 12 LAM Partners had made data publicly available.

With the rapid digital transformation of cultural heritage institutions, virtual guides have emerged as an innovative solution to enhance visitor engagement in museums. However, the effectiveness of these digital guides might depend not only on the technology itself but also on how well they match the artifacts. Grounded in the diffusion of innovations theory, this study investigates how the perceived cultural similarity between a digital guide’s appearance and the cultural style of exhibited artifacts influences visitors’ experience in museums. We conducted a single-factor two-level between-subjects laboratory experiment. The results indicate a mediating role for perceived compatibility with the moderation of openness trait. In addition, visitor experience was found to predict purchase intention. These findings offer new insights into how digital innovations can be more effectively diffused in cultural settings by emphasizing the role of perceived compatibility and providing direction in designing culturally consistent virtual guides to improve visitor experience.