With the goal of developing information retrieval (IR) systems, including digital libraries (DLs), that support universal access, the authors conducted two studies to identify the help-seeking situations that blind and visually impaired (BVI) users encountered in DLs. First, the authors quantitatively compared BVI users’ help-seeking situations when interacting with the DL's mobile website (M.Web) and the mobile app (M.App). Using multiple data collection methods, the mobile study identified six situations BVI users faced more frequently when using M.Web than M.App. Second, the findings of the mobile study were qualitatively compared with the situations that occurred in the desktop environment, highlighting eight unique situations. It is more difficult for BVI users to interact with DLs in the mobile environment because of the reduced size of the mobile interface, complex DL structures, dynamic elements, diverse formats, and sight-centered design. Specific design recommendations were offered to enhance DLs in the mobile environment.

Diffusion-based generative models have achieved remarkable success in image generation. Their guidance formulation allows an external model to plug-and-play control the generation process for various tasks without finetuning the diffusion model. However, the direct use of publicly available off-the-shelf models for guidance fails due to their poor performance on noisy inputs. For that, the existing practice is to fine-tune the guidance models with labeled data corrupted with noises. In this paper, we argue that this practice has limitations in two aspects: (1) performing on inputs with extremely various noises is too hard for a single guidance model; (2) collecting labeled datasets hinders scaling up for various tasks. To tackle the limitations, we propose a novel strategy that leverages multiple experts where each expert is specialized in a particular noise range and guides the reverse process of the diffusion at its corresponding timesteps. However, as it is infeasible to manage multiple networks and utilize labeled data, we present a practical guidance framework termed Practical Plug-And-Play (PPAP), which leverages parameter-efficient fine-tuning and data-free knowledge transfer. We exhaustively conduct ImageNet class conditional generation experiments to show that our method can successfully guide diffusion with small trainable parameters and no labeled data. Finally, we show that image classifiers, depth estimators, and semantic segmentation models can guide publicly available GLIDE through our framework in a plug-and-play manner. Our code is available at https://github.com/riiid/PPAP.

Abstract This is the first study that compares types of orientation tactics that blind and sighted users applied in their initial interactions with a digital library (DL) and the associated factors. Multiple methods were employed for data collection: questionnaires, think‐aloud protocols, and transaction logs. The paper identifies seven types of orientation tactics applied by the two groups of users. While sighted users focused on skimming DL content, blind users concentrated on exploring DL structure. Moreover, the authors discovered 13 types of system, user, and interaction factors that led to the use of orientation tactics. More system factors than user factors affect blind users' tactics in browsing DL structures. The findings of this study support the social model that the sight‐centered design of DLs, rather than blind users' disability, prohibits them from effectively interacting with a DL. Simultaneously, the results reveal the limitation of existing interactive information retrieval models that do not take people with disabilities into consideration. DL design implications are discussed based on the identified factors.