through Cultural Engagement) project. The dataset serves as a tool for identifying and examining communities that emerge from citizens' interactions with cultural heritage, capturing key representations of individuals and groups to reveal unexpected connections. An example from a SPICE case study at GAM (Galleria Civica d'Arte Moderna e Contemporanea) in Turin, Italy, illustrates the dataset's structure, focusing on the interpretation of artworks, with particular attention to the deaf community's emotional responses. The dataset primarily organizes diverse perspectives, values, and emotions expressed through non-transitive relationships. Initially designed for analyzing narrative identities within museum audiences and communities, the dataset has potential applications in education, social work, and community building.
Italy ∗Corresponding author. †These authors contributed equally.
CEUR
ceur-ws.org pretation and reflection are closely interconnected: the project introduces a continuous Interpretation Reflection Loop (IRL), which serves as a model for linking interpretation and reflection activities across diferent phases and components of the SPICE digital platform. The SPICE platform provides users with various citizen curation activities, such as selecting artifacts, tagging, and sharing personal stories and opinions. Through user-friendly interfaces, citizens are encouraged to contribute and share their rich interpretations of the cultural heritage artifacts they encounter. These contributions are then analyzed by the system to promote reflection. The platform may suggest new activities, ofer recommendations, and present users with alternative perspectives, enabling them to explore and reflect on both their own contributions and those of others. This creates an inclusive and participatory interpretation reflection loop.
In this paper, we describe a dataset collected during the conduction of the SPICE case study carried
out it Turin from May 2020 to April 2023 in cooperation with Galleria Civica d’Arte Moderna e
Contemporanea, GAM 1), in which we developed a web app for citizen curation called GAMGame.
The GAMGame, inspired by theories of “cultural narrative identity” [
Using the GAMGame, users can create and share stories from the museum artworks in a simple intuitive. To create a story, the user select and added the artworks from the artwork gallery; when an artwork is added, the applications requires the user to annotate it with “something personal” using hashtags, emojis and short text templates. When using the app, the users can receive afect-based recommendations about artworks and stories which are aimed at expanding diversity in story creation and exploration.
The dataset collected during the testing of the GAMGame has been integrated into the VISIR software
tool, which helps identify and explore the communities emerging from citizens’ interactions with
cultural heritage through a visual interface [
The dataset includes the following attributes: • Citizen demographic attributes: for citizens, 5 diferent attributes were collected from the users of the GAMGame through online forms: relationshsip_with_arts encodes the citizen’s interest in art; relationship_with_museums encodes the frequency with which the citizen visit museums; the binary attribute interest_in_LIS encodes the interest in contents in Italian Sign Language (Lingua Italiana dei Segni, Italian Sign Language) (LIS); gender (male, female and non-binary) and age range are standard demographic variables. • Artworks attributes: this set of attributes includes both standard museum catalogue metadata, such as title, technique, dimension (size_height, size_width, size_depth), collection, year (including artwork_start_date and artwork_end_date) and type, and a set of attributes added by the museum staf to investigate the visitors’ relationship with the artworks. In particular, based on the experience gathered during museum labs with schools and communities, the latter include information on the artwork’s subject, materials, and artistic movement (artwork_artistic_movement), since the visitors tend to engage with the subject depicted by the artwork, the artistic movement it belongs to, and to the materials that compose it; since visitors are sometimes aware of the artist’s main biographic data, specific attributes for the artist’s nationality ( artist_country), birth and death date
Age range 0,0% 4,2%
13,1%
(artist_birth_date, artist_death_date), and gender were also included. To avoid arbitrariness in
identifying the subject values, we resorted to an authoritative, standard resource, the Iconclass2
classification of iconographic subjects. Iconographic subjects are organized into 9 main subject
types (Abstract, Non-representational Art; Religion and Magic; Nature; Human Being, Man in
General; Society, Civilization, Culture; Abstract Ideas and Concepts; History; Bible; Literature;
Classical Mythology and Ancient History), further subdivided into more specific types of varying
specificity (e.g., “adolescent, young woman, maiden” or “sea (seascape)”, recurring subjects in the
GAM collection).
• Interpretation attributes: for each artwork, a list of emotion labels was extracted from the
curatorial notes, and from the stories generated by the users through the GAMGame. Curatorial
notes and user-generated text in English and Italian (comments and tags) were fed into the
semantic analysis and reasoning pipeline of SPICE, which relies on Plutchik’s theory of emotions
[
The enrichment of the catalogue records with the attributes described above (Artwork attributes) was conducted on a set 56 items included in the GAMGame installation created for the pilot. The rational behind the selection consisted in presenting the users with a variety of artworks by author, type, style and time period, so as to avoid any biases towards specific artwork types. This choice is reflected by the value distributions of the artwork attributes, which encompass 12 diferent types of material, 26 artistic movements, 21 techniques, 4 artwork types and artists from 10 diferent countries.
During the European Researchers’ Night at University of Torino in 2022, SPICE researchers collected 149 stories from 49 casual users who volunteered to use the GAMGame app. Figure 1 shows the pie chart statistics on the user attributes: users are evenly distributed according to gender (about 43.8% of male and female)(Figure 1 (b)); most of them are aged between 20 and 30 years old (about 58.3%) (Figure 1 (a)); they have a strong interest in receiving contents in Italian Sign Language (Lingua Italiana dei Segni, Italian Sign Language (LIS)) (52.1%) (Figure 1 (c)) and in art (64.6%) (Figure 1 (e)). Finally, 37.7% users stated that they often visit museums and art galleries (Figure 1 (d)). Concerning the stories, they contained 56 distinct artworks, i.e., all artworks have been selected by the users at least once. The stories contained overall 404 artworks, which yields an average story length of 2.7 artworks.
The dataset has been designed and implemented to be managed by a NoSQL Database Management System (DBMS), stored and distributed across multiple servers within a Linked Data Hub (LDH) system. This approach ensures a more flexible data model compared to relational databases, allowing for greater scalability in service utilization, higher availability and fault tolerance. The dataset consists of three JSON files: • artworks - This dataset contains detailed records of the artworks and includes metadata such as titles, authors, creation dates, dimensions, materials, techniques, and collections they belong to. Each entry also includes descriptions, artistic movements, associated emotions, and other relevant classifications. • ugcUsers - This dataset contains information about users who contribute content related to the artworks. This may include their profiles, activities, contributions, and their interactions with the artworks in the database. It can be used to track the engagement of diferent users with the cultural heritage content. • ugcContributions - This dataset contains the contributions made by GAMGame users. These include comments, emotions, discussions, and other forms of user-generated content associated with the artworks. This file links the users (from ugcUsers.json) with their contributions, providing insights into how communities interact with and contribute to the database. An example of this entry is described in Figure 2.
Together, these datasets provide a snapshot of the interaction between users and cultural heritage in citizen curation, allowing for a deeper understanding of the engagement with the artworks through cultural narratives.
The first dataset in JSON format named “artworks.json” contains detailed records of artworks, primarily from a cultural heritage of GAM’s collection. Each entry in this dataset includes various attributes related to individual artworks. Here’s a breakdown of the key components found in the file: 1. Basic Information: 2. Artwork Details: 3. Descriptive Information: • ID: A unique identifier for each artwork. • Title: The title of the artwork in GAM’s Museum. • Inventory Number: A unique code identifying the artwork within the collection. • Collection: The specific collection or exhibition the artwork is part of. • Author: The artist who created the artwork. • Year: The year the artwork was created. • Material and Technique: The materials used and the technique applied to create the artwork. • Dimensions: The physical dimensions of the artwork, including height, width, and depth. • Acquisition: Information on how and when the artwork was acquired by the collection. • Description: A narrative description providing context, historical significance, or a detailed analysis of the artwork. • Image URL: A link to an image of the artwork.
• Artistic Movement: The art movement or style to which the artwork belongs. 4. Artist Information: • Birth and Death Dates: The lifespan of the artist. • Country: The nationality of the artist.
• Gender: The gender of the artist.
• Extracted Emotions: Emotions associated with or evoked by the artwork, as identified by curators or through our system DEGARI 2.0 for emotions extraction. • Iconclass Subjects: Iconographic subjects and themes depicted in the artwork, classified according to the
Iconclass system (ICONCLASS API3 for artworks) • Artwork Type: The type of artwork, such as painting, sculpture, etc. • Technique: Specific techniques used in the creation of the artwork.
• Materials: The materials used in the artwork.
The second JSON dataset (‘ugcUsers.json‘) contains user demographic data related to their interactions with a cultural heritage or art-related application. Here’s a summary of the key elements found in the dataset: 1. User Identifiers : • Each user in the dataset is represented by a unique ‘userid‘. This identifier is used to group the diferent pieces of demographic data related to a specific user. 2. Demographic Information: The dataset includes several demographic attributes for each user, such as: 3. Context and Category: 4. Data Structure: • Gender: Recorded as values like “Male “Female” or ”Not specified”. • Age: Age ranges such as “20-30 age”, ”31-45 age”, ”45-60 age”, etc. • Relationship With Art: Indicates the user’s interest in art, with values like “I have a strong interest in art”, “Art interests me little”, or “I have no interest in art”, • Relationship With Museum: Describes the user’s frequency of museum visits, with values like ”I often visit museums”, “I visit museums and exhibitions from time to time”, or “I rarely visit museums and exhibitions. • Content in LIS (Italian Sign Language): Indicates the user’s interest in content provided in Sign Language, such as “I’m not interested in sign language content” or ”I would like to see content in Sign Language (LIS)”. • Deaf: A binary value (“Yes” or “No”) indicating whether the user is deaf. • Each demographic entry is associated with a ‘context‘, typically set to ”application”, which likely refers to the context in which this data was collected. • The ‘category‘ for all entries is ”demographics,” highlighting that this dataset focuses on user demographic information. • The data is structured with each ‘userid‘ being a key that maps to an array of demographic entries. Each entry in this array includes fields like ‘id‘, ‘source_id‘, ‘source‘, ‘pname‘ (parameter name), ‘pvalue‘ (parameter value), ‘context‘, and ‘datapoints‘.
This dataset can be used to analyze user demographics and their relationship with art and museums, as well as their preferences for accessible content like LIS. This information could be valuable for tailoring cultural heritage experiences or exhibitions to diferent audience segments.
degari extracted emotions
Array of Strings
id DateOfLastModify title Inventary Collection Sumbject author year Material_and_echnique Dimension Definizione Acquisizione image description .
Artwork start date Artwork end date Artist birth date Artist death date Gender Artist country Artist secondary country Artwork Artistic Movement Secondary Artwork Artistic Movement Technique Artwork type Size unity Size height Size width Size depth Materials Iconclass subjects curators iconclassIDString
A unique identifier for each artwork entry in the GAM’s collection.
The date and time when the record was last modified. The title of the artwork.
The inventory number of the artwork within the GAM’s collection.
The specific collection or exhibition the artwork is part of.
The subject of the artwork (often left blank or empty). The name of the artist who created the artwork. The year when the artwork was created.
The materials and techniques used in the creation of the artwork.
The physical dimensions of the artwork, including height, width, and depth.
Additional definitions or classifications of the artwork (often left blank).
Information on how and when the artwork was acquired by the collection.
A URL linking to an image of the artwork.
A narrative description providing context, historical significance, or a detailed analysis of the artwork. Emotions associated with or evoked by the artwork, as identified by curators or through analysis.
The start date of the artwork’s creation.
The end date of the artwork’s creation (often left blank). The birth date of the artist.
The death date of the artist (if applicable).
The gender of the artist.
The country of origin or nationality of the artist. A secondary country associated with the artist (if applicable) The primary artistic movement or style to which the artwork belongs.
A secondary artistic movement or style associated with the artwork (if applicable).
Specific techniques used in the creation of the artwork. The type of artwork, such as painting, sculpture, etc. The unit of measurement for the artwork’s dimensions (e.g., cm).
The height of the artwork.
The width of the artwork.
The depth of the artwork (if applicable).
A list of materials used in the artwork.
Iconographic subjects and themes depicted in the artwork, classified according to the Iconclass system. A string representing the Iconclass IDs associated with the artwork’s themes and subjects.
An array of Iconclass IDs linked to the artwork. The approximate year of the artwork’s creation. The decade during which the artwork was created.
Field Name userid id
A unique identifier for each user. It is used to track the contributions made by individual users.
The ID uniquely identifies a story and is automatically generated by the GAMGame app each time a new story is created. In Figure 2, user with userid BTKF72et created the story with is 632dab3fcfb08e51f124bcef, which can contain a minimum of 2 and a maximum of 3 artworks selected by the user (each artwork is uniquely identified by the ”origin” attribute conteined in the dataset ”artworks.json” shown in Figure ??). Represents the source or context of the emotional response, likely linking back to a specific artwork selected by the user. In particular, in Figure 2, the user with userid BTKF72et selected the painting with id 39347 (Dans mon pays by Marc Chagall4), generating the emotion Sadness with an intensity of 0.93.
An identifier linking the response to the source of the stimulus, possibly an internal reference within the application.
Indicates the origin of the data, often marked as “fake” in this dataset, suggesting placeholder or anonymized data.
The parameter name, typically set to itMakesMeThinkAbout.emotions, (Used to describe the emotions evoked by a memory when admiring the painting) or itRemindsMeOf.emotions (indicating to describe the emotions evoked by the type of memories that the painting brings to mind) and finally itMakesMeFeel.emotions (used to convey the emotions that are expressed and the feelings experienced when admiring that particular painting).
Contains the actual emotional response data, represented as a dictionary with keys being emotion names (e.g., sadness, joy) and values being the intensity of these emotions.
Describes the context in which the data was collected, usually set to “application”, indicating it was gathered through the GAMGame app. Numerical value representing the aggregation level of the data, often set to 0 in this dataset.
Indicates the type of data, commonly set to “interest”, reflecting the user’s engagement or emotional response to the content.
A unique identifier for each user. It is used to associate demographic data and preferences with individual users.
A unique identifier for each demographic entry associated with a user.
Indicates the origin of the data, often set as “User”, indicating that the data was input by the user. An identifier linking the demographic entry to its source within the application.
Indicates the origin of the demographic data, often labeled as “fake”, suggesting placeholder or anonymized data.
The parameter name describing the demographic aspect being recorded, such as Gender, Age, RelationshipWithArt, etc.
Contains the actual demographic data, with values corresponding to the parameter name, such as “Male” for Gender, “20-30 age” for Age, etc.
Describes the context in which the demographic data was collected, typically set to “application”.
Numerical value representing the aggregation level of the data, often set to 0 in this dataset.
Indicates the category of data, which in this dataset is consistently “demographics”, reflecting that the entries are related to user demographic information.
The research leading to this publication has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreements ”SPICE - Social Cohesion, Participation, and Inclusion through Cultural Engagement” (Grant Agreement N. 870811), https://spice-h2020.eu.
From this GitHub repository, you can download our dataset in JSON-LDH format. Although the dataset currently lacks machine readable semantic metadata, the vocabulary illustrated in this paper for data description is aligned with the SPICE Ontology Network (SON)5 in compliance with the FAIR paradigm.
• GitHub