Post-doctorate reseracher Maija Spuriņa spent the first three months of the project working with literature on digitalization of memory at the atvian National LIbrary and the British Library in London. 

From bibliography covered so far the most significant have been monogrtaphies: 

• Lev Manovich “Languages of the new media” (2001), 
• Jacque Derrida “Archive Fever” (1996), 
• David Weinberger “Everything is Miscellaneous: The Power of the New Digital Disorder” (2007), 
• Manuel Castells “The Rise of Network Society” (2010), 
• Alison Lansberg “Prosthetic Memory” (2004). 
• Jose Van Dijck “Mediated Memories in the Digital Age” (2007) 
• Susan M. Pearce “Museums, Objects and Collections: A Cultural Study” (1993). 

Also the following collections of articles have been covered: 

• Andrew Hoskins (ed) “Digital Memory Studies: Media Pasts in Transition” (2017). 
• Joanne Garde-Hansen, Andrew Hoskins, Anna Reading (ed) “Save As …. Digital memories” (2009). 
• Astrid Erll, Ann Rigney (ed) “Mediation, remediation, and the dynamics of cultural memory“ (2009). 
• Susan A. Crane (ed) “Museums and Memory” (2000). 
• Ross Parry (ed) “Museums in a Digital Age” (2010). 
• Neiger, M., O. Meyers, E. Zandberg (ed) “On Media Memory: Collective Memory in a New Media Age” (2011). 

Four books from those mentioned above have been bought with the funds of the project (Van Dijck (2007), Pearce (1993), Crane (2000), Ross (2010)) and will be available at the library of the academy. 

The postdoctoral researcher presented the project in February in the seminar series “Tuesdays of New Academics”. 

17.04.2019. 

From June 24 to 29, Maija Spuriņa, a researcher from the Scientific Research Center of the Latvian Academy of Culture (LKA), participated in the annual conference of the Memory Studies Association, where she presented her project on the digitalization and transnationalization of memory in a panel discussion titled Rethinking Archives of Memory. She also participated in a total of nine panel discussions, including topics on memory digitalization, global memories, and memory and populism.   

The topic of memory digitalization was discussed in several panels. Innovative methods were presented, showing how various digital technologies can be used to collect data that can be utilized in social process research. For example, a separate panel was devoted to sensorial memories, where several researchers shared their methods for collecting sensor data, including using sensors to record physiological body reactions. 

The panel where Maija Spuriņa presented her research was focused on memory archives. The panel included researchers from Brazil, the Philippines, and the United Kingdom. Two studies examined digital oral history archives. The discussion raised questions about the use, authorship, and ethical aspects of these memories. One of the most valuable insights in the context of the digitalization process in memory institutions was the realization that we are in a transition phase between analog and digital order, often using digital tools based on analog logic. During this transition phase, the view of digitalization is frequently either utopian (associated with excessive optimism that digital will solve all problems) or dystopian (associated with fears of the unknown and worries that the digital will destroy the existing world order). 

From a practical perspective, the panel discussion on publishing opportunities in the field of memory studies was very valuable. The discussion included editors from both memory studies journals—the Memory Studies Journal and History and Memory—along with representatives from several publishers who publish special series dedicated to memory studies: Palgrave Macmillan Memory Studies Series, Sage Worlds of Memory Series, and DeGruyter Cultural Memory and Media Series. All publishers invited proposals for monographs and conceptual collections of articles. 

From July 22 to August 3, Maija Spuriņa participated in the European Summer University in Digital Humanities Culture & Technology program, organized by Leipzig University for the tenth consecutive year. Out of the twelve workshops offered, which included 18 sessions in small groups (5-10 people), Maija Spuriņa took part in the workshop Images of Image Machines. Theory and Practice of Interpretable Machine Learning for the Digital Humanities. In addition to the workshops, she attended lectures on the development of digital humanities in different parts of the world, including Russia, Brazil, Cameroon, Germany, and Europe in general. She also visited three workshops on database modeling, XML, and humanities data, and explored colleagues' projects in various fields of digital humanities. 

During the summer school, Maija Spuriņa not only enhanced her skills in working with the Python programming language, which she had already learned before going to Leipzig through the Massachusetts Institute of Technology (MIT) online course platform Edx, but also gained a deeper understanding of machine learning algorithms in general, their principles, applications, possibilities and limitations, and most importantly, their social consequences regarding development and usage. 

Digital Humanities (DH) is a field where the humanities intersect with computer science. Looking at the overall development of DH, two significant questions can be distinguished. First: What can computer technology and computer science provide to the humanities and social sciences? And second: What can the humanities and social sciences offer to computer science and technology development? So far, the key focus in the development of DH has been on the first question. DH training programs, summer schools, and conferences have mainly focused on teaching humanities and social scientists how to process quantitative data and use digital tools for the needs of the humanities. However, in recent years, it has increasingly been acknowledged at IT and DH conferences that it is important to consider social and cultural aspects in the development of IT tools. In recent years, the hope and belief that digital data analysis will provide a more objective view of social reality have been diminished by research on social bias, which, for various reasons, is reflected in the obtained results. The perspective of the social and humanities sciences is crucial for critically evaluating IT-developed tools and algorithms and reducing the hidden social biases within them. 

(Maija Spuriņa's participation in the summer university was supported by the ERDF postdoctoral fellowship program, project no. 1.1.1.2/VIAA/2/18/252). 

The Latvian Academy of Culture was represented at the European Sociological Association conference in Manchester, which took place from August 20 to 23, by Anda Laķe, Agnese Hermane, Maija Spuriņa, and Iveta Ķešāne. In the cultural sociology session on digitalization, Maija Spuriņa presented her project on memory digitalization and transnationalization (project no. 1.1.1.2/VIAA/2/18/252). The presentation was highly successful, with particular interest in the application of machine learning algorithms in image analysis used in the research. The postdoctoral researcher established connections with colleagues from Portugal, Italy, the Netherlands, and the United States, which could potentially develop into valuable collaborative projects in the future. 

Digital technologies, their social aspects of development, and their impact on society were explored in many conference sessions and presentations. One of the most interesting discussions focused on algorithms and their societal influence. Eran Fisher from Israel presented a study on the navigation app Waze and its impact on local communities. Italian sociologist Lorenza Bonina discussed the use of algorithms in education and related challenges, while Polish academic Jernej Prodnik analyzed the connection between algorithms and the acceleration of social life in journalism. Another compelling presentation was by Jamie Woodcock from the University of Oxford, who spoke about the work experiences of digital platform employees and their ability to mobilize, organize, and advocate for their rights. His observations were primarily based on interviews and fieldwork with Uber drivers (similar to Latvia’s Bolt) and food delivery workers (similar to Latvia’s Wolt). Woodcock concluded that platform workers are socially vulnerable, subject to economic exploitation, and have fewer opportunities for political mobilization compared to factory workers of the industrial era. However, he also identified several factors that have so far contributed to platform workers’ mobilization and cooperation. 

The dominant themes of the conference were immigration, nationalism, and populism. Among all the semi-plenary discussions, the most inspiring were the speeches by Elijah Anderson, Michèle Lamont, and Ruth Wodak. Anderson, a professor of sociology and African American studies at Yale University, is one of the most well-known urban ethnographers in the U.S. and Europe, focusing primarily on race in urban settings. In Manchester, he spoke about how a Black person experiences life in a "white people’s world." His observations were deeply personal, ethnographically rich, and emotionally moving, sparking reflections on the genealogy of social stereotypes and prejudices. Professor Ruth Wodak from Lancaster University analyzed the rhetoric of right-wing populists, primarily through examples from Austrian and U.S. political discourse. Michèle Lamont, on the other hand, addressed the crisis of "hope narratives." The "American Dream" as the dominant hope narrative in the U.S. has lost its legitimacy, as it no longer provides a meaningful and attainable vision of the future for any social class—including the economically privileged, who suffer from infertility and depression. Lamont called for the creation of new hope narratives based on inclusivity, pluralism, everyday universalism, and cosmopolitanism.   

06.09.2019. 

The first year of the research project Digitalization of Social Memory: Digitalization Practices in National Museums and Their Impact on Museum Nationalization and Transnationalization has been successfully completed. During this period, postdoctoral researcher Maija Spuriņa not only familiarized herself with the relevant literature and conducted a preliminary study but also learned the basics of the Python programming language and digital humanities methods. She has shared this knowledge through several lectures at the Latvian Academy of Culture, the Art Academy of Latvia, and an open lecture within the Cultural Crossroads conference, as well as by presenting at international conferences in Madrid, Leipzig, and Manchester. 

A successful collaboration has been established with the National Museum of Art and the Cultural Information Systems Centre (KISC). The postdoctoral researcher has started working with data prepared by KISC from the National Museum Joint Catalogue (NMKK). Using the data processing tool OpenRefine, Maija has begun working with publicly available NMKK data, which consists of more than 400 fields and over 6 million records. 

07.01.2020. 

In the first half of the second year of the research project, the postdoctoral researcher continued analyzing the digitized resources of Latvian museums available to the public on the National Museum Collection Joint Catalogue (NMKK) website nmkk.lv. The project initially planned to present the analysis results at the Nordic Digital Humanities Conference in Riga, scheduled for March 2020. Unfortunately, due to the COVID-19 pandemic, the conference was postponed to October of the same year. Below is a summary of the theses the postdoctoral researcher had hoped to present at the canceled conference. 

Thanks to digitalization, data on more than 1.3 million objects stored in Latvian museums are accessible to everyone, including over 270,000 images. In a way, this information can be considered the digital representation of Latvia’s national museum collection, making it crucial, from the perspective of social memory, to understand what kind of perception of Latvia’s cultural heritage it creates.   

However, like many large datasets, the NMKK data is not easily navigable or intuitively comprehensible. Currently, users can “wander” through digitized museum collections, exploring objects’ titles, dimensions, materials, descriptions, and, in some cases, images. From a cultural research perspective, however, a more valuable approach would be to gain an overview of what is stored in Latvian museums. For example, it would be insightful to determine which historical periods or geographic regions are represented, what types of objects are available, or what materials are most common. It would also be interesting to examine specific categories of objects, such as vases or crowns, analyzing their forms and color diversity. As media theorist Lev Manovich points out, in theory, the digitization of cultural heritage should allow for both qualitative and quantitative research into cultural artifacts—for example, understanding what people photographed in 1930s Europe or even globally. However, Manovich concludes that currently, digitized cultural resources do not provide such an overview. Instead, they resemble isolated cultural islands, making it difficult to compile or compare them. 

The same applies to the National Museum Collection Joint Catalogue (NMKK) of Latvia. At present, the catalog consists of separate islands of information, created by various museums, which are difficult to compile and analyze comprehensively. One of the postdoctoral researcher’s key goals in this phase of the project was to apply data analysis methods to gain a clearer overview of the NMKK dataset and to identify the obstacles preventing such an overview. Using Python, OpenRefine, and Excel, the researcher analyzed the materials, object types, origins, and time periods of digitized items in Latvian museums. One of the most challenging tasks was compiling information on the historical periods represented in NMKK. The existing data formats for object dating are so diverse that they cannot be consolidated effectively.   

Alongside NMKK data analysis, the postdoctoral researcher also identified similar databases in neighboring countries—Estonia and Finland—for comparison in the later stages of the project. In Finland, this is FINNA, a cultural resource content database that includes digitized museum collections. In Estonia, this is MUIS, the national museum database. The researcher established contacts with colleagues at the University of Tartu and the National Library of Finland, aiming to compare all three databases in the upcoming stages of the project. 

In this stage of the project the researcher continued analysis of the data available on nmkk.lv as well as started inquiry into digital colection of Finnland available on FINNA.fi and Estonia – MUIS.ee.

The researcher took par tin two conferences – an international conference “Culture Crossroads” and Annual Conference of Nordic Digital Humanities. Both conference were hosted in Riga and took place via ZOOM. The researcher presented findings of her project focusing on the potential of and barriers to data aggregation in digital museum collections. Both presentations are available on academi.edu.

Even though digitization of cultural heritage provides an opportunity to aggregate a huge amount of data and get a real burd-view on national and even trans-national cultural resources, in reality there are numerous digital barriers that hinder such aggregation. For example, Latvian museum database NMKK contains information on more than 1,3 million objects.  Its infrustructure constitutes 438 information spaces for each objects, thus providing a potential space for more than 600 million units of information.  In reallyty only about 31 million spaces are filled with actual information and a vast majority of this information cannot be aggregated due to discrepancies in formats. Only 20 fields from, the total of 438, are used by multiple museums. The fragmentation of data can be explained by several factors. First, museums themselves and the objects they store are very diverse and cannot be described in uniform terms. For example, the museums of natural history hold objects that do not have “Date of Production”, “Date of Use” or an “Author”, while majority of other museums use these fields to describe their collections. Second, each museum has its own tradition of organizing and describing its collections. Finally, data agrggation system in Latvia so far has been vey democratic – museums have been allowed to use any fields upon their choice and even to introduce new fields to describe their collections. The resulting data base is a collection of huge amount of information with almost no structure and with very weak possibilities of aggregation and overview.

The researcher analysed possibilities to aggregate NMKK data using several sociologically significant parameters, such as time, space, material, object type, and image. She consluded that the two, from the sociologial view point probably most significant, parameters – time and space – cannot  be used for aggregation.  Information on date of object is found across 47 fiels, information on space – across 13 fields.  Even if compiled, the information on time appear in vastly diverse formats that do not allow computerized processings. Some of the formats used are as follows:

19th century

19th century 50ies

19th century 50ies – 70ies

18th – 19th century AC

the end of 19th century (middle, beginning)

the last quarter of 19the century

around 1900

01.01.1899.

01.01.1899.- 01.01.1899.

January 1899

etc.

Even though this information is intelligable from a human point of view and can be manually processed, it cannot be processed via algoritm.

The researcher also started inquiry in to FINNA.fi – got familiar with the structure of the databse, the documentation and potential ways to access it via API.  She learned webscraping with API using information available and consulkting with collegues at the National Library of Finnland. She also developed a Python algorythm for webscraping images available in FINNA and started data downloading process.

In the first half of 2021, work on preparing three datasets was completed, and the analysis and comparison of each dataset began. The data includes images and metadata of objects stored in museums in Latvia, Finland, and Estonia, as available in the databases nmkk.lv, finna.fi, and muis.ee. For the analysis, all available images of physical objects from these databases were used, totaling approximately 900,000 images. These were obtained through web scraping and are being analyzed using the image recognition machine learning algorithm VGG19. Additionally, metadata for all NMKK, FINNA, and MUIS objects is being utilized.  At this stage, an analysis of the manufacturing dates of objects has been conducted, providing insights into the "memory density" of different historical periods. This required extensive data cleaning and recoding. The materials of the digitized objects were also analyzed as part of the study. 

In the second quarter of the third year of the project, the density of cultural memory was analyzed using data on tangible objects from the digitized museum collection databases NMKK, FINNA, and MUIS. 

The results are presented in graphical form here: 

Figure 1. Memory density: Latvia, Estonia, Finland (1 AD – 2020 AD). Number of tangible objects by date (weighted). Total number of data entries: 740,000. 

Figure 2. Memory density: Latvia, Estonia, Finland (1800 – 2020). Number of tangible objects by date (weighted). Total number of data entries: 740,000. 
 

The CVGG19 machine learning algorithm was also used to gain a transnational view of specific historical periods. Here is an example of the representation of the 19th century in the data of tangible objects stored and digitized in the museum collections of Latvia and Finland. 

Figure 3. The 19th century "machine vision" of tangible objects stored and digitized in the museum collections of Latvia and Finland. Total number of data entries: 376,010. 

During this period, the postdoctoral researcher participated in three scientific conferences. At the annual conference of the Memory Studies Association in Warsaw, she presented a paper titled "Digital boundaries and convergences of digitized museum collection databases – Latvian NMKK and Finnish FINNA" in the panel Digital Memory: Museums. She also presented at the annual conference of the European Sociological Association in Barcelona with a paper titled Islands and Ocean of Digitized Cultural Memory and ‘Machine Vision’. 

The key points of the presentation were as follows: 

• The memory theorists Jan and Aleida Assmann distinguish two parts of cultural memory – functional memory and archive memory. Functional memory is a small part of the past that is actively used, politically significant, and closely related to the present. The majority of cultural memory is archive memory, stored in the archives of museums, libraries, and archives, and not actively used. 
• Archive memory plays a crucial role – it serves as a reservoir for alternative functional memories. It is a source of critique for politically engaged functional memory and enables the potential change of functional memories. 
• Archive memory consists of millions of objects (documents, books, artworks, tangible objects) located in countless storage facilities. Therefore, it is difficult to subject it to empirical research. 
• Thanks to the digitalization of archives, it is now possible to browse and analyze archive memory.  
• Using data from national digitized museum databases, one can analyze the density of museum archive memory, i.e., how extensive the collections of various historical periods are in the archives. The NMKK and FINNA data show that the most densely represented periods are the 19th and 20th centuries. 
• Thanks to digitalization, everyday users can now wander through countless museum collections. However, the most significant potential benefit of digitalizing museum collections is the possibility of combining, reviewing, and analyzing these collections as a unified dataset, potentially revealing new connections and insights. 
• Digitized museum objects usually consist of a digital image and the object's metadata, which informs about the object’s name, author, creation time, location, material, and other characteristics that vary depending on the type of object and the museum that created the data. 
• For digitized data to form a single entity – metaphorically, a digital ocean of data – there must be touchpoints between digitized objects that allow them to be aggregated, compared, and analyzed. Paradoxically, at present, the easiest way to link collections from different museums is through images and processing them using machine learning algorithms for semantic recognition. The use of metadata is hindered by language barriers and differing data formats and structures. 
• It can be concluded that, at present, digitized museum collections resemble digital islands of data, which are difficult to integrate into a unified digital ocean due to the significant amount of data cleaning, translation, and interpretation required. 

In the second half of 2021, the postdoctoral researcher prepared a scientific publication titled Shape of Storage Memory: A Digital Analysis of the Museum Storages of Northeast Europe, which was submitted to the leading memory studies journal Memory Studies, published by SAGE Publishing. The pre-publication of the article is available on ResearchGate: https://www.researchgate.net/publication/356784278_Shape_of_Storage_Memory_A_Digital_Analysis_of_the_Museum_Storages_of_Northeast_Europe. 

Here is the abstract of the article in English: 

Although storage memory serves an important social function, its dispersion across various institutional storerooms, as well as the diversity of shapes and formats of objects, hinders its analysis. Mass digitization of museum collections provides an opportunity to analyze storage memory and its shape and structure. It transcends national boundaries as it pools together digitized data from multiple countries. In this article, I present the transnational mnemonic density of Northeast European storage memory through an analysis of digital databases of museum collections from Latvia, Estonia, and Finland. Using metadata on time, I illuminate the overall shape of Northeast European storage, identify and interpret socio-mnemonically dense and scarce periods, and reflect on the difficulties encountered during the data analysis process, which are informative about the nature of digitized storage memory. 

Additionally, the postdoctoral researcher prepared a second scientific article titled “The Semantic Gap between Human and Machine in Databases of Digitised Museum Collections,” which was submitted to one of the most prestigious cultural studies journals – Media, Culture & Society (SAGE). 

Here is the abstract of the article in English: 

Digitization of museum collections transforms material traces of the pre-digital past, diverse in materials and shapes, to a digital format. This transformation requires creative translation of each cultural artifact from a format that is accessible, informative, and meaningful for a human being to one readable by a machine. The paper addresses differences between these two formats through a critical analysis of data in two national databases of digitized museum collections – FINNA of Finland and NMKK of Latvia. By exploring obstacles for data aggregation and formatting inconsistencies, it sheds light on the semantic gap between the human and machine view of the world. 

Finally, during this period, the postdoctoral researcher participated in a panel discussion organized by the Latvian Cultural Information Systems Centre. The discussion was streamed online, and its recording is available for all interested parties: https://www.kis.gov.lv/lv/jaunums/krajuma-digitalizacijas-darba-svariga-kvalitate-un-ilgtspeja#tiessaistes-seminara-muzeju-krajuma-digitalizacija-ieraksts.