Keywords

1 Introduction and Theoretical Background

The research Project “Development of Innovative Applications through the Exploitation of Landmarks for the Promotion of Ancient Greek Technology Exhibits.”—project code: Τ2ΕΔΚ-03,578 which co-financed by the European Regional Development Fund of the European Union and Greek national funds through the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call RESEARCH—CREATE—INNOVATE, aimed to expand technological boundaries in cultural tourism using augmented reality and narration techniques. The integrated narrative creation platform utilises a focused documentary format to provide historical exhibits with a spatial connection to the surrounding area. The platform’s comprehensive approach covers multiple types of experience. It offers interactive narration through augmented reality (AR), which is a successful approach as seen in many studies. It utilizes interactive narration through augmented reality, which has been demonstrated to be effective in various implementations [1,2,3]. Creating a fun user experience with Augmented Reality (AR) is also an important factor [4] which supports participation in active learning processes [5, 6]. Enhancing the user’s experiential walk through storytelling and rich material and displaying augmented reality cultural content [7, 8] according to personalised user requirements using machine learning techniques [9] completes the experience. The scientific study of ancient Greek technology was documented by Kotsanas after years of research and publication [10].

The app displays 3D visuals of ancient Greek technological inventions and augmented reality simulations of their use. MAET also created a documentation format for exhibits and components. Each section was linked to more than two landmarks (GPS) or beacons in 3D prints for indoor storytelling, as in the exhibitions. 3D photorealistic models of the exhibits were saved on a platform with interactive simulations of their functions. The 3D models were optimised for mobile devices, particularly for AR. User preferences were collected through gamification and matched with ancient inventions to create a personal narrative. Tourists, educators, students, and MAET visitors evaluated this project. The narrative can lead users to visit exhibitions and participate in museum activities. Users’ feedback was vital and was gathered through SUS questionnaires completed by those who visited the exhibition spaces after using the applications during the pilot period. The anonymous data collected from the applications through the Ambient Intelligence Environment, once reaching a satisfactory volume, will support future machine learning mechanisms facilitated by the platform. This research resulted in an intelligent environment that supports augmented reality applications in the cultural heritage domain. This environment featured an innovative model that ensured sustainability and user engagement; technical diagnostics deeply tied to the environment for assessing resilience and human-induced environmental changes relative to the cultural product; and a knowledge management framework that could gather, manage, and disseminate knowledge related to cultural heritage and associated activities.

2 Methodology

2.1 Ambient Intelligence Environment

To develop a digital platform for creating and managing stories on mobile devices (PCMS), MAET has developed appropriate documentation formats for exhibits and their main components. Each section was associated with more than two points on the map (GPS coordinates), connected to a specific ontology of concepts, or to the context of indoor storytelling (periodic exhibitions, exhibition stands, etc.) with specific beacons integrated into 3D printed parts. All parts of the exhibits were stored on the platform in 3D photorealistic models, and there was a fully interactive simulation of their operation, except for the 3D model of the exhibit itself. The 3D models introduced on the platform were designed to operate on mobile devices for use in augmented reality applications. To create the necessary 3D exhibits and their 3D simulation, MAET, as a manufacturer of fully functional exhibits of ancient Greek technology, and in collaboration with Omega Technology, an IT company with extensive experience in 3D digital modelling, produced the required 3D products. The architecture selected for the platform development is shown in Fig. 2 at Kotsopoulos et al. [11].

The platform for creating and managing narration has the following capabilities regarding exhibits: brief description, functions, list of exhibits, classification of exhibits, and creation of exhibits (exhibit registration form: Name, Description, category, edit of exhibit, addition of three-dimensional, two-dimensional, photographs, audio to exhibit). The exhibit sections have the following registration management possibilities: brief description, functions, list of sections, classification of sections, and creation of exhibit (registration form exhibit section: name, photo, display of exhibit, addition of three-dimensional, two-dimensional, photographs, and audio to an exhibit section). In addition, the rewards of users are defined based on their achievements, connection with beacons, registration of locations and their connection to the departments of exhibits, and all other functionalities that support the mobile application.

2.2 Personalized Recommendation System

The personalised recommendation system proposes a list of locations considering the user’s profile, interests, history of interactions with other locations, and location characteristics. For the creation of the personalized proposal system, the Collaborative Filtering (CF) technique was used. In its general version, this technique can provide suggestions and predictions regarding a user’s choices, based on the choice history and preferences of multiple users of the same system. The basic idea of this technique is that if two people have similar preferences/tastes/ideas on one topic, there is a possibility that their views will be identical. In this case, users with similar characteristics are likely to choose to visit related sites of archaeological interest.

The creation of a forecasting system based on collaborative filtering requires a large dataset with site ratings and user characteristics. Owing to the lack of a free dataset that can be used to train the model, it was decided that in the first phase and for the trial period, the system proposals would emerge in a separate manner. The criterion by which a service is recommended to a user is the similarity metric between the user and site. The similarity metric chosen is the Jaccard similarity index which calculates the similarity of two sets A, B. In our case, set A consisted of the preference tags selected by each user and set B consisted of the tags of each location.

In this phase, based on the user’s geographical location, the list of locations closest to the user is calculated, and for each location, the value mentioned in the previous paragraph is calculated. The list of sites is sorted based on this user-site similarity value. Sites with a high degree of similarity to the user are higher on the list. The sorted list is a system recommendation system for the users. However, training a CF model requires the creation of a model-training dataset. The dataset will be gradually built during the system’s use by real users of the application. To create the set, each time the user visits an area, they are asked to rate the site with a score between 1 and 5 at a later time. It has been observed that many users of electronic systems are not willing to evaluate products and services they use. Therefore, we decided to provide extra incentives to users to rate sites by offering them several points. Note that the points are exchangeable for products from the application’s virtual store.

To enhance the quality of forecasts, to continue using the extra information we have about user characteristics and to address the problem of CF, Cold Start systems (poor forecast quality for new users whose evaluation history does not exist) and the CMFREC library were used [12]. This library offers the possibility of implementing a hybrid CF model that uses user ratings and additional information regarding users and locations as inputs (Figs. 1, 2, 3, 4).

Fig. 1
A screenshot from a mobile application. It has two photos under the popular destinations based on your preferences category, one of a road and another of a theater. Additionally, there is a photo of a site under the routes for you category. The text is in a foreign language.

Screens from the mobile application

Full size image
Fig. 2
A screenshot of a mobile application page. The text at the top reads exhibits. It has four photos of 4 different objects. The text under them is in a foreign language.

Screens from the mobile application

Full size image
Fig. 3
A screenshot of a mobile application. It has two photos of sites under the Nearby Locations category. There are 3 thumbnail photos of objects at the top with text in a foreign language.

Screens from the mobile application

Full size image
Fig. 4
A screenshot of a mobile app. It has a fragment augmented reality view of an object. There is a card with a thumbnail of an object at the bottom. The text is in a foreign language.

Screens from the mobile application

Full size image

3 Results and Discussion

3.1 Evaluation Period Results

Initially, the application was tested by MAET museum visitors. After a small presentation of the application, visitors of the MAET museums were suggested to download the application, test the application, and answer a small questionnaire after the testing. Among the users who downloaded and tested the app, 42 visitors from Athens answered the online questionnaire and 18 visitors from the Ancient Olympia. In addition, it was tested using the beacon version of the app embedded in exhibits at the 2023 Ecsite Conference by sixteen visitors.

All users answered questionnaires using the System Usability Scale (SUS) [13]. The results of SUS were evaluated according to Sauro [14] based on their mean score, interpreting it into grades (A, B, C, D, F), adjectives (best imaginable, Excellent, Good, Ok, Poor, Worst imaginable), acceptability (Acceptable, Marginal, Not Acceptable), and NPS categories (Promoter, Passive, Detractor). For the 42 users of Athens with a mean SUS score of 80.3, the result is good, acceptable, and graded as A, and 17 users can be characterised as promoters and 25 users as passive. For the 18 users of the Ancient Olympia with a mean SUS score of 78.5, the result is acceptable, graded as A; seven users can be characterised as promoters, nine users as passive, and two users as detractors. For the 16 users of the 2023 Ecsite Conference with a mean SUS score of 80.0, the result is acceptable, graded as A, and six users can be characterised as promoters and 10 users as passive. Together, the present findings confirm that the users of the app who tested it during the evaluation period reacted positively to its use. One concern regarding the findings of the evaluation period was that the users who tested the app were informed by the staff of the MAET Museum. Thus, the users did not reach MAET Museums because of the app, which will be the next target after uploading the app to app stores and conducting appropriate marketing to promote it.

3.2 Suggested Ambient Intelligence Approach

The central core of ambient intelligence was the proposal of an appropriate narrative that would satisfy the user at the level of landmarks he will visit, at the level of interest in the exhibits to visit the MAET, and finally at the level of general personal interests. Specifically, the implemented approach successfully correlates:

The types of visitors to MAET Museums: The study of the visitor profile utilises a detailed analysis of the data recorded by the Kotsanas Museum in Athens and Ancient Olympia over the last three (3) years. Demographic data, educational level, social dimension, country of origin, nationality, culture, educational or other placement, individual visit, or group are the products of study to achieve results in relation to the development of the narrative scenario. The types of Gamification Users: (Social (wants to socialise), Autonomous (wants complete freedom of movement and expression), Altruist (wants to help others), Skillful (wants to achieve goals), Player (wants to earn rewards)) for mobile applications (utilizing features such as cooperation, obstacles, goals, communication, imagination, strategy, and exploration). The Landmarks: Landmarks are concerned with the categorisation of visiting and selection points, which directly and indirectly determine the final decision of a visitor. The extent to which a destination is desirable, the uniqueness of the spectacle presented [15], the acceptance that a place enjoys through public reviews or recommendations, as well as personalised prioritisation combined with location, “out of the box” logic and specific topics, will be recorded through selected tags and spatial associations. The Exhibits: Exhibits (e.g., themes, level of interaction and functionality, scale/size, exhibits aimed at younger or adult audiences) related to the type of visitors of MAET museums. The interviews with MAET executives contributed significantly to the research on the initial categorisation of visitors based on the exhibits they preferred to learn, the time they devoted to each of them, the level of absorption and reproduction of information to third parties, and the preferred actions of MAET (thematic visit experiences, special tours, workshops, educational programs, multi-thematic actions, and open events).

A similar study by Ramos et al. [16] endeavoured to auto-curate tourist pathways tailored to individual and group visitor profiles, considering factors such as mood, personality, and context, underpinned by an Ambient Intelligence framework, but it is not included in the context exhibits from a museum. Piccialli et al. [17] delved into harnessing data science, specifically machine learning, to valorise and promote cultural heritage, focusing on analysing IoT-derived visitor data from the National Archaeological Museum of Naples, but there is no connection with the landmarks of Naples. In a recent study by Casillo et al. [18], the integration of context-awareness techniques into Recommender Systems in the era of Big Data, emphasising their transformative impact on enhancing recommendations in the Cultural Heritage sector, is examined, pointing to its importance.

The need to create some basic rules regarding narrative suggestions to users of the AR application, how they will be displayed, how they relate to the user’s behaviour, and possible preferences regarding exhibits were initially approached using content-based recommendation techniques. Contextual filtering approaches only consider an individual user’s previous preferences and attempt to train a preference model based on a preference representation of the content of the proposed items. An important part of the suggestion model process is the additional features of the proposals, such as special attributes, descriptions, metadata, comments, and the user’s definition of preferences [19]. After a period when the application will be used, a sufficient user base will be created with their preferences imprinted. Each user and element were described by a feature vector. In this way, users and elements are integrated into a common space; thus, studying the behaviours of other users will propose appropriate narratives to a user using collaborative filtering techniques.

4 Conclusion

In this study, we introduced an innovative approach to interlinking landmarks with simulated exhibits, considering the unique characteristics of various users, from tourists at destinations to museum visitors. This pioneering research delves into a potentially unparalleled domain, emphasising that the foundation for establishing guidelines leading to tailored user proposals can only be truly validated through hands-on applications by end-users during advanced implementation stages. Our primary objective was to design an advanced environment conducive to augmented reality applications, specifically for the cultural heritage sector. The key features of this study include a groundbreaking framework that not only ensures the sustainability of these applications, but also enhances user enjoyment, aligning with organizational benefits. An efficient knowledge-management system is skilled in collecting, managing, and disseminating information about cultural heritage and human engagement. The resources generated, coupled with a structured digital knowledge repository, stand as unmatched assets, elevating the possibility of creating distinctive and influential applications. Our approach was designed to amplify museum goals by drawing more visitors and offering immersive experiences that commence externally and deepen internally. Simultaneously, it seeks to elevate visitors’ appreciation of the city’s cultural fabric and heighten their enthusiasm for global exhibits, expanding both the hosting venues and audience reach. In summary, this research has the potential to benefit a wide range of stakeholders, from cultural institutions and tourists to educators, researchers, and local businesses. Further research and experiments on user experience of mobile applications that focus on understanding its features can increase user acceptance.