A Methodology for Semantic Enrichment of Cultural Heritage Images Using Artificial Intelligence Technologies

Abstract

Cultural heritage images are among the primary media for communicating and preserving the cultural values of a society. The images represent concrete and abstract content and symbolise the social, economic, political, and cultural values of the society. However, an enormous amount of such values embedded in the images is left unexploited partly due to the absence of methodological and technical solutions to capture, represent, and exploit the latent information. With the emergence of new technologies and availability of cultural heritage images in digital formats, the methodology followed to semantically enrich and utilise such resources become a vital factor in supporting users need. This paper presents a methodology proposed to unearth the cultural information communicated via cultural digital images by applying Artificial Intelligence (AI) technologies (such as Computer Vision (CV) and semantic web technologies). To this end, the paper presents a methodology that enables efficient analysis and enrichment of a large collection of cultural images covering all the major phases and tasks. The proposed method is applied and tested using a case study on cultural image collections from the Europeana platform. The paper further presents the analysis of the case study, the challenges, the lessons learned, and promising future research areas on the topic.

Hypericons for Abstract Concepts: Non-traditional Approaches to Visual Explainability

“…Specifically, ground-truth generation is highly inconsistent, caused by personal and situational factors such as cultural background, personality, and social context [59]. Almost all trainable classes in popular datasets for image classification, such as ImageNet and Google Open Images, refer to concrete classes [60,61]. Moreover, abstract words tend to have higher dispersion ratings due to the wide variety of images returned from a query [62,63].…”

“…Some work has attempted to grapple with the problems inherent in evaluating subjective tasks in the context of computer vision. In cultural image labeling with abstract concepts, [60] proposes fuzzy annotation to account for uncertainty [86,87]. For visual sentiment analysis, [68] suggests multi-label learning with unequal label importance, predicting a probability distribution of emotions.…”

“…Moreover, abstract words tend to have higher dispersion ratings due to the wide variety of images returned from a query [62,63]. Despite these issues, a few works define their task as grappling with abstract concept detection from images [64][65][66]. For example, [60] propose a framework to label cultural food images with one abstract concept, while [66] builds on previous work [65] focusing on classifying advertisement images based on symbol clusters.…”

Evaluation Method of Financial Internal Control in Colleges and Universities Based on Artificial Intelligence Technology

“…As shown in Figure 3, we can obtain the output results of the evaluation data of the college financial internal control, realize the data collection process of the evaluation of the college financial internal control, and on this basis, access the report data in the offline state, compare whether there is an error in the data point, if not, continue to process, and if so, output the location information of the point as the result of locating the error data [22,23]. For the characteristic data with obvious correlation, the relevant data can be combined into one piece of data by operating the data, and other data can be deleted.…”

Editorial for Special Issue “Fine Art Pattern Extraction and Recognition”

“…The results obtained suggest that it is possible to generate iconographically significant captions that capture not only the objects depicted, but also the historical and artistic context of an artwork. Abgaz et al [11] also focused on the semantic enrichment of digitized cultural images and introduce a methodology for fully exploiting latent cultural information that is communicated visually by applying a combination of computer vision and semantic web technologies. A case study on food images is presented.…”