Author Archives: tpadilla

Call for Proposals: Advancing Inclusive Computational Research with Archives Research Compute Hub

Last summer, Internet Archive launched ARCH (Archives Research Compute Hub), a research service that supports creation, computational analysis, sharing, and preservation of research datasets from terabytes and even petabytes of data from digital collections – with an initial focus on web archive collections. In line with Internet Archive’s mission to provide “universal access to all knowledge” we aim to make ARCH as universally accessible as possible. 

Computational research and education cannot remain solely accessible to the world’s most well-resourced organizations.  With philanthropic support, Internet Archive is initiating Advancing Inclusive Computational Research with ARCH, a pilot program specifically designed to support an initial cohort of five less well-resourced organizations throughout the world. 

Opportunity

  • Organizational access to ARCH for 1 year – supporting research teams, pedagogical efforts, and/or library, archive, and museum worker experimentation.  
  • Access to thousands of curated web archive collections – abundant thematic range with potential to drive multidisciplinary research and education. 
  • Enhanced Internet Archive training and support – expert synchronous and asynchronous support from Internet Archive staff. 
  • Cohort experience – opportunities to share challenges and successes with a supportive group of peers. 

Eligibility

  • Demonstrated need-based rationale for participation in Advancing Inclusive Computational Research with Archives Research Compute Hub: we will take a number of factors into consideration, including but not limited to stated organizational resources relative to peer organizations, ongoing experience contending with historic and contemporary inequities, as well as levels of national development as assessed by the United Nations Least Developed Countries effort and Human Development Index
  • Organization type: universities, research institutes, libraries, archives, museums, government offices, non-governmental organizations. 

Apply

Submission deadline: 2/26/2024

Decisions communicated to applications: 3/11/2024

Program begins: 3/25/2024

Apply here. 

Wrapping up Legal Literacies for Text and Data Mining – Cross-Border (LLTDM-X)

In August 2022, the UC Berkeley Library and Internet Archive were awarded a grant from the National Endowment for the Humanities (NEH) to study legal and ethical issues in cross-border text and data mining (TDM).

The project, entitled Legal Literacies for Text Data Mining – Cross-Border (“LLTDM-X”), supported research and analysis to address law and policy issues faced by U.S. digital humanities practitioners whose text data mining research and practice intersects with foreign-held or licensed content, or involves international research collaborations.

LLTDM-X is now complete, resulting in the publication of an instructive case study for researchers and white paper. Both resources are explained in greater detail below.

Project Origins

LLTDM-X built upon the previous NEH-sponsored institute, Building Legal Literacies for Text Data Mining. That institute provided training, guidance, and strategies to digital humanities TDM researchers on navigating legal literacies for text data mining (including copyright, contracts, privacy, and ethics) within a U.S. context.

A common challenge highlighted during the institute was the fact that TDM practitioners encounter expanding and increasingly complex cross-border legal problems. These include situations in which: (i) the materials they want to mine are housed in a foreign jurisdiction, or are otherwise subject to foreign database licensing or laws; (ii) the human subjects they are studying or who created the underlying content reside in another country; or, (iii) the colleagues with whom they are collaborating reside abroad, yielding uncertainty about which country’s laws, agreements, and policies apply.

Project Design

LLTDM-X was designed to identify and better understand the cross-border issues that digital humanities TDM practitioners face, with the aim of using these issues to inform prospective research and education. Secondarily, it was hoped that LLTDM-X would also suggest preliminary guidance to include in future educational materials. In early 2023, the project hosted a series of three online round tables with U.S.-based cross-border TDM practitioners and law and ethics experts from six countries. 

The round table conversations were structured to illustrate the empirical issues that researchers face, and also for the practitioners to benefit from preliminary advice on legal and ethical challenges. Upon the completion of the round tables, the LLTDM-X project team created a hypothetical case study that (i) reflects the observed cross-border LLTDM issues and (ii) contains preliminary analysis to facilitate the development of future instructional materials.

The project team also charged the experts with providing responsive and tailored written feedback to the practitioners about how they might address specific cross-border issues relevant to each of their projects.

Guidance & Analysis

Case Study

Extrapolating from the issues analyzed in the round tables, the practitioners’ statements, and the experts’ written analyses, the Project Team developed a hypothetical case study reflective of “typical” cross-border LLTDM issues that U.S.-based practitioners encounter. The case study provides basic guidance to support U.S. researchers in navigating cross-border TDM issues, while also highlighting questions that would benefit from further research. 

The case study examines cross-border copyright, contracts, and privacy & ethics variables across two distinct paradigms: first, a situation where U.S.-based researchers perform all TDM acts in the U.S., and second, a situation where U.S.-based researchers engage with collaborators abroad, or otherwise perform TDM acts in both U.S. and abroad.

White Paper

The LLTDM-X white paper provides a comprehensive description of the project, including origins and goals, contributors, activities, and outcomes. Of particular note are several project takeaways and recommendations, which the project team hopes will help inform future research and action to support cross-border text data mining. Project takeaways touched on seven key themes: 

  1. Uncertainty about cross-border LLTDM issues indeed hinders U.S. TDM researchers, confirming the need for education about cross-border legal issues; 
  2. The expansion of education regarding U.S. LLTDM literacies remains essential, and should continue in parallel to cross-border education; 
  3. Disparities in national copyright, contracts, and privacy laws may incentivize TDM researcher “forum shopping” and exacerbate research bias;
  4. License agreements (and the concept of “contractual override”) often dominate the overall analysis of cross-border TDM permissibility;
  5. Emerging lawsuits about generative artificial intelligence may impact future understanding of fair use and other research exceptions; 
  6. Research is needed into issues of foreign jurisdiction, likelihood of lawsuits in foreign countries, and likelihood of enforcement of foreign judgments in the U.S. However, the overall “risk” of proceeding with cross-border TDM research may remain difficult to quantify; and
  7. Institutional review boards (IRBs) have an opportunity to explore a new role or build partnerships to support researchers engaged in cross-border TDM.

Gratitude & Next Steps

Thank you to the practitioners, experts, project team, and generous funding of the National Endowment for the Humanities for making this project a success. 

We aim to broadly share our project outputs to continue helping U.S.-based TDM researchers navigate cross-border LLTDM hurdles. We will continue to speak publicly to educate researchers and the TDM community regarding project takeaways, and to advocate for legal and ethical experts to undertake the essential research questions and begin developing much-needed educational materials. And, we will continue to encourage the integration of LLTDM literacies into digital humanities curricula, to facilitate both domestic and cross-border TDM research.

[Note: this content is cross-posted on the Legal Literacies for Text and Data Mining project site and the UC Berkeley Library Update blog.]

Leveraging Technology to Scale Library Research Support: ARCH, AI, and the Humanities

Kevin Hegg is Head of Digital Projects at James Madison University Libraries (JMU). Kevin has held many technology positions within JMU Libraries. His experience spans a wide variety of technology work, from managing computer labs and server hardware to developing a large open-source software initiative. We are thankful to Kevin for taking time to talk with us about his experience with ARCH (Archives Research Compute Hub), AI, and supporting research at JMU

Thomas Padilla is Deputy Director, Archiving and Data Services. 

Thomas: Thank you for agreeing to talk more about your experience with ARCH, AI, and supporting research. I find that folks are often curious about what set of interests and experiences prepares someone to work in these areas. Can you tell us a bit about yourself and how you began doing this kind of work?

Kevin: Over the span of 27 years, I have held several technology roles within James Madison University (JMU) Libraries. My experience ranges from managing computer labs and server hardware to developing a large open-source software initiative adopted by numerous academic universities across the world. Today I manage a small team that supports faculty and students as they design, implement, and evaluate digital projects that enhance, transform, and promote scholarship, teaching, and learning. I also co-manage Histories Along the Blue Ridge which hosts over 50,000 digitized legal documents from courthouses along Virginia’s Blue Ridge mountains.

Thomas: I gather that your initial interest in using ARCH was to see what potential it afforded for working with James Madison University’s Mapping Black Digital and Public Humanities project. Can you introduce the project to our readers? 

Kevin: The Mapping the Black Digital and Public Humanities project began at JMU in Fall 2022. The project draws inspiration from established resources such as the Colored Convention Project and the Reviews in Digital Humanities journal. It employs Airtable for data collection and Tableau for data visualization. The website features a map that not only geographically locates over 440 Black digital and public humanities projects across the United States but also offers detailed information about each initiative. The project is a collaborative endeavor involving JMU graduate students and faculty, in close alliance with JMU Libraries. Over the past year, this interdisciplinary team has dedicated hundreds of hours to data collection, data visualization, and website development.

Mapping the Black Digital and Public Humanities, project and organization type distribution

The project has achieved significant milestones. In Fall 2022, Mollie Godfrey and Seán McCarthy, the project leaders, authored, “Race, Space, and Celebrating Simms: Mapping Strategies for Black Feminist Biographical Recovery“, highlighting the value of such mapping projects. At the same time, graduate student Iliana Cosme-Brooks undertook a monumental data collection effort. During the winter months, Mollie and Seán spearheaded an effort to refine the categories and terms used in the project through comprehensive research and user testing. By Spring 2023, the project was integrated into the academic curriculum, where a class of graduate students actively contributed to its inaugural phase. Funding was obtained to maintain and update the database and map during the summer.

Looking ahead, the project team plans to present their work at academic conferences and aims to diversify the team’s expertise further. The overarching objective is to enhance the visibility and interconnectedness of Black digital and public humanities projects, while also welcoming external contributions for the initiative’s continual refinement and expansion.

Thomas: It sounds like the project adopts a holistic approach to experimenting with and integrating the functionality of a wide range of tools and methods (e.g., mapping, data visualization). How do you see tools like ARCH fitting into the project and research services more broadly? What tools and methods have you used in combination with ARCH?

Kevin: ARCH offers faculty and students an invaluable resource for digital scholarship by providing expansive, high-quality datasets. These datasets enable more sophisticated data analytics than typically encountered in undergraduate pedagogy, revealing patterns and trends that would otherwise remain obscured. Despite the increasing importance of digital humanities, a significant portion of faculty and students lack advanced coding skills. The advent of AI-assisted coding platforms like ChatGPT and GitHub CoPilot has democratized access to programming languages such as Python and JavaScript, facilitating their integration into academic research.

For my work, I employed ChatGPT and CoPilot to further process ARCH datasets derived from a curated sample of 20 websites focused on Black digital and public humanities. Utilizing PyCharm—an IDE freely available for educational purposes—and the CoPilot extension, my coding efficiency improved tenfold.

Next, I leveraged ChatGPT’s Advanced Data Analysis plugin to deconstruct visualizations from Stanford’s Palladio platform, a tool commonly used for exploratory data visualizations but lacking a means for sharing the visualizations. With the aid of ChatGPT, I developed JavaScript-based web applications that faithfully replicate Palladio’s graph and gallery visualizations. Specifically, I instructed ChatGPT to employ the D3 JavaScript library for ingesting my modified ARCH datasets into client-side web applications. The final products, including HTML, JavaScript, and CSV files, were made publicly accessible via GitHub Pages (see my graph and gallery on GitHub Pages)

Black Digital and Public Humanities websites, graph visualization

In summary, the integration of Python and AI-assisted coding tools has not only enhanced my use of ARCH datasets but also enabled the creation of client-side web applications for data visualization.

Thomas: Beyond pairing ChatGPT with ARCH, what additional uses are you anticipating for AI-driven tools in your work?

Kevin: AI-driven tools have already radically transformed my daily work. I am using AI to reduce or even eliminate repetitive, mindless tasks that take tens or hundreds of hours. For example, as part of the Mapping project, ChatGPT+ helped me transform an AirTable with almost 500 rows and two dozen columns into a series of 500 blog posts on a WordPress site. ChatGPT+ understands the structure of a WordPress export file. After a couple of hours of iterating through my design requirements with ChatGPT, I was able to import 500 blog posts into a WordPress website. Without this intervention, this task would have required over a hundred hours of tedious copying and pasting. Additionally, we have been using AI-enabled platforms like Otter and Descript to transcribe oral interviews.

I foresee AI-driven tools playing an increasingly pivotal role in many facets of my work. For instance, natural language processing could automate the categorization and summarization of large text-based datasets, making archival research more efficient and our analyses richer. AI can also be used to identify entities in large archival datasets. Archives hold a treasure trove of artifacts waiting to be described and discovered. AI offers tools that will supercharge our construction of finding aids and item-level metadata.  

Lastly, AI could facilitate more dynamic and interactive data visualizations, like the ones I published on GitHub Pages. These will offer users a more engaging experience when interacting with our research findings. Overall, the potential of AI is vast, and I’m excited to integrate more AI-driven tools into JMU’s classrooms and research ecosystem.

Thomas: Thanks for taking the time Kevin. To close out, whose work would you like people to know more about? 

Kevin: Engaging in Digital Humanities (DH) within the academic library setting is a distinct privilege, one that requires a collaborative ethos. I am fortunate to be a member of an exceptional team at JMU Libraries, a collective too expansive to fully acknowledge here. AI has introduced transformative tools that border on magic. However, loosely paraphrasing Immanuel Kant, it’s crucial to remember that technology devoid of content is empty. I will use this opportunity to spotlight the contributions of three JMU faculty whose work celebrates our local community and furthers social justice.

Mollie Godfrey (Department of English) and Seán McCarthy (Writing, Rhetoric, and Technical Communication) are the visionaries behind two inspiring initiatives: the Mapping Project and the Celebrating Simms Project. The latter serves as a digital, post-custodial archive honoring Lucy F. Simms, an educator born into enslavement in 1856 who impacted three generations of young students in our local community. Both Godfrey and McCarthy have cultivated deep, lasting connections within Harrisonburg’s Black community. Their work strikes a balance between celebration and reparation. Collaborating with them has been as rewarding as it is challenging.

Gianluca De Fazio (Justice Studies) spearheads the Racial Terror: Lynching in Virginia project, illuminating a grim chapter of Virginia’s past. His relentless dedication led to the installation of a historical marker commemorating the tragic lynching of Charlotte Harris. De Fazio, along with colleagues, has also developed nine lesson plans based on this research, which are now integrated into high school curricula. My collaboration with him was a catalyst for pursuing a master’s degree in American History.

Racial Terror: Lynching in Virginia

Both the Celebrating Simms and Racial Terror projects are highlighted in the Mapping the Black Digital and Public Humanities initiative. The privilege of contributing to such impactful projects alongside such dedicated individuals has rendered my extensive tenure at JMU both meaningful and, I hope, enduring.

Build, Access, Analyze: Introducing ARCH (Archives Research Compute Hub)

We are excited to announce the public availability of ARCH (Archives Research Compute Hub), a new research and education service that helps users easily build, access, and analyze digital collections computationally at scale. ARCH represents a combination of the Internet Archive’s experience supporting computational research for more than a decade by providing large-scale data to researchers and dataset-oriented service integrations like ARS (Archive-it Research Services) and a collaboration with the Archives Unleashed project of the University of Waterloo and York University. Development of ARCH was generously supported by the Mellon Foundation.

ARCH Dashboard

What does ARCH do?

ARCH helps users easily conduct and support computational research with digital collections at scale – e.g., text and data mining, data science, digital scholarship, machine learning, and more. Users can build custom research collections relevant to a wide range of subjects, generate and access research-ready datasets from collections, and analyze those datasets. In line with best practices in reproducibility, ARCH supports open publication and preservation of user-generated datasets. ARCH is currently optimized for working with tens of thousands of web archive collections, covering a broad range of subjects, events, and timeframes, and the platform is actively expanding to include digitized text and image collections. ARCH also works with various portions of the overall Wayback Machine global web archive totaling 50+ PB going back to 1996, representing an extensive archive of contemporary history and communication.

ARCH, In-Browser Visualization

Who is ARCH for? 

ARCH is for any user that seeks an accessible approach to working with digital collections computationally at scale. Possible users include but are not limited to researchers exploring disciplinary questions, educators seeking to foster computational methods in the classroom, journalists tracking changes in web-based communication over time, to librarians and archivists seeking to support the development of computational literacies across disciplines. Recent research efforts making use of ARCH include but are not limited to analysis of COVID-19 crisis communications, health misinformation, Latin American women’s rights movements, and post-conflict societies during reconciliation. 

ARCH, Generate Datasets

What are core ARCH features?

Build: Leverage ARCH capabilities to build custom research collections that are well scoped for specific research and education purposes.

Access: Generate more than a dozen different research-ready datasets (e.g., full text, images, pdfs, graph data, and more) from digital collections with the click of a button. Download generated datasets directly in-browser or via API. 

Analyze: Easily work with research-ready datasets in interactive computational environments and applications like Jupyter Notebooks, Google CoLab, Gephi, and Voyant and produce in-browser visualizations.

Publish and Preserve: Openly publish datasets in line with best practices in reproducible research. All published datasets will be preserved in perpetuity. 

Support: Make use of synchronous and asynchronous technical support, online trainings, and extensive help center documentation.

How can I learn more about ARCH?

To learn more about ARCH please reach out via the following form

Collective Web-Based Art Preservation and Access at Scale 

Art historians, critics, curators, humanities scholars and many others rely on the records of artists, galleries, museums, and arts organizations to conduct historical research and to understand and contextualize contemporary artistic practice. Yet, much of the art-related materials that were once published in print form are now available primarily or solely on the web and are thus ephemeral by nature. In response to this challenge, more than 40 art libraries spent the last 3 years developing a collective approach to preservation of web-based art materials at scale. 

Supported by the Institute of Museum and Library Services and the National Endowment for the Humanities, The Collaborative ART Archive (CARTA) community has successfully aligned effort across libraries large and small, from Manoa, Hawaii to Toronto, Ontario and back resulting in preservation of and access to 800 web-based art resources, organized into 8 collections (art criticism, art fairs and events, art galleries, art history and scholarship, artists websites, arts education, arts organizations, auction houses), totalling nearly 9 TBs of data with continued growth. All collections are preserved in perpetuity by the Internet Archive. 

Today, CARTA is excited to launch the CARTA portal – providing unified access to CARTA collections.

CARTA portal

🎨 CARTA portal 🎨

The CARTA portal includes web archive collections developed jointly by CARTA members, as well as preexisting art-related collections from CARTA institutions, and non-CARTA member collections. CARTA portal development builds on the Internet Archive’s experience creating the COVID-19 Web Archive and Community Webs portal. 

CARTA collections are searchable by contributing organization, collection, site, and page text. Advanced search supports more granular exploration by host, results per host, file types, and beginning and end dates.

CARTA search

🔭 CARTA search 🔭

In addition to the CARTA portal, CARTA has worked to promote research use of collections through a series of day long computational research workshops – Working to Advance Library Support for Web Archive Researchbacked by ARCH (Archives Research Compute Hub). A call for applications for the next workshop, held concurrent to the annual Society of American Archivists meeting, is now open. 

Moving forward CARTA aims to grow and diversify its membership in order to increase collective ability to preserve web-based art materials. If your art library would like to join CARTA please express interest here..

Getting Started with Machine Learning and GLAM (Galleries, Libraries, Archives, Museums) Collections

Guest Post by Daniel Van Strien, Machine Learning Librarian, Hugging Face

Machine learning has many potential applications for working with GLAM (galleries, libraries, archives, museums) collections, though it is not always clear how to get started. This post outlines some of the possible ways in which open source machine learning tools from the Hugging Face ecosystem can be used to explore web archive collections made available via the Internet Archive’s ARCH (Archives Research Compute Hub). ARCH aims to make computational work with web archives more accessible by streamlining web archive data access, visualization, analysis, and sharing. Hugging Face is focused on the democratization of good machine learning. A key component of this is not only making models available but also doing extensive work around the ethical use of machine learning. 

Below, I work with the Collaborative Art Archive (CARTA) collection focused on artist websites. This post is accompanied by an ARCH Image Dataset Explorer Demo. The goal of this post is to show how using a specific set of open source machine learning models can help you explore a large dataset through image search, image classification, and model training. 

Later this year, Internet Archive and Hugging Face will organize a hands-on hackathon focused on using open source machine learning tools with web archives. Please let us know if you are interested in participating by filling out this form.

Choosing machine learning models

The Hugging Face Hub is a central repository which provides access to open source machine learning models, datasets and demos. Currently, the Hugging Face Hub has over 150,000 openly available machine learning models covering a broad range of machine learning tasks.

Rather than relying on a single model that may not be comprehensive enough, we’ll select a series of models that suit our particular needs.

A screenshot of the Hugging Face hub task navigator presenting a way of filtering machine learning models hosted on the hub by the tasks they intend to solve. Example tasks are Image Classification, Token Classification and Image-to-Text.

A screenshot of the Hugging Face Hub task navigator presenting a way of filtering machine learning models hosted on the hub by the tasks they intend to solve. Example tasks are Image Classification, Token Classification and Image-to-Text.

Working with image data 

ARCH currently provides access to 16 different “research ready” datasets generated from web archive collections. These include but are not limited to datasets containing all extracted text from the web pages in a collection, link graphs (showing how websites link to other websites), and named entities (for example, mentions of people and places). One of the datasets is made available as a CSV file, containing information about the images from webpages in the collection, including when the image was collected, when the live image was last modified, a URL for the image, and a filename.

Screenshot of the ARCH interface showing a preview for a dataset. This preview includes a download link and an “Open in Colab” button.

Screenshot of the ARCH interface showing a preview for a dataset. This preview includes a download link and an “Open in Colab” button.

One of the challenges we face with a collection like this is being able to work at a larger scale to understand what is contained within it – looking through 1000s of images is going to be challenging. We address that challenge by making use of tools that help us better understand a collection at scale.

Building a user interface

Gradio is an open source library supported by Hugging Face that helps create user interfaces that allow other people to interact with various aspects of a machine learning system, including the datasets and models. I used Gradio in combination with Spaces to make an application publicly available within minutes, without having to set up and manage a server or hosting. See the docs for more information on using Spaces. Below, I show examples of using Gradio as an interface for applying machine learning tools to ARCH generated data.

Exploring images

I use the Gradio tab for random images to begin assessing images in the dataset. Looking at a randomized grid of images gives a better idea of what kind of images are in the dataset. This begins to give us a sense of what is represented in the collection (e.g., art, objects, people, etc.).

Screenshot of the random image gallery showing a grid of images from the dataset.

Screenshot of the random image gallery showing a grid of images from the dataset.

Introducing image search models 

Looking at snapshots of the collection gives us a starting point for exploring what kinds of images are included in the collection. We can augment our approach by implementing image search. 

There are various approaches we could take which would allow us to search our images. If we have the text surrounding an image, we could use this as a proxy for what the image might contain. For example, we might assume that if the text next to an image contains the words “a picture of my dog snowy”, then the image contains a picture of a dog. This approach has limitations – text might be missing, unrelated or only capture a small part of what is in an image. The text “a picture of my dog snowy” doesn’t tell us what kind of dog the image contains or if other things are included in that photo.

Making use of an embedding model offers another path forward. Embeddings essentially take an input i.e. text or image, and return a bunch of numbers. For example, the text prompt: ‘an image of a dog’, would be passed through an embedding model, which ‘translates’ text into a matrix of numbers (essentially a grid of numbers). What is special about these numbers is that they should capture some semantic information about the input; the embedding for a picture of a dog should somehow capture the fact that there is a dog in the image. Since these embeddings consist of numbers, we can also compare one embedding to another to see how close they are to each other. We expect the embeddings for similar images to be closer to each other and the embeddings for images which are less similar to each other to be farther away. Without getting too much into the weeds of how this works, it’s worth mentioning that these embeddings don’t just represent one aspect of an image, i.e. the main object it contains but also other components, such as its aesthetic style. You can find a longer explanation of how this works in this post.

Finding a suitable image search model on the Hugging Face Hub

To create an image search system for the dataset, we need a model to create embeddings. Fortunately, the Hugging Face Hub makes it easy to find models for this.  

The Hub has various models that support building an image search system. 

A screenshot of the Hugging Face Hub showing a list of hosted models.

Hugging Face Hub showing a list of hosted models.

All models will have various benefits and tradeoffs. For example, some models will be much larger. This can make a model more accurate but also make it harder to run on standard computer hardware.

Hugging Face Hub provides an ‘inference widget’, which allows interactive exploration of a model to see what sort of output it provides. This can be very useful for quickly understanding whether a model will be helpful or not. 

A screenshot of a model widget showing a picture of a dog and a cat playing the guitar. The widget assigns the label `"playing music`" the highest confidence.

A screenshot of a model widget showing a picture of a dog and a cat playing the guitar. The widget assigns the label `”playing music`” the highest confidence.

For our use case, we need a model which allows us to embed both our input text, for example, “an image of a dog,” and compare that to embeddings for all the images in our dataset to see which are the closest matches. We use a variant of the CLIP model hosted on Hugging Face Hub: clip-ViT-B-16. This allows us to turn both our text and images into embeddings and return the images which most closely match our text prompt.

A screenshot of the search tab showing a search for “landscape photograph” in a text box and a grid of images resulting from the search. This includes two images containing trees and images containing the sky and clouds.

Aa screenshot of the search tab showing a search for “landscape photograph” in a text box and a grid of images resulting from the search. This includes two images containing trees and images containing the sky and clouds.

While the search implementation isn’t perfect, it does give us an additional entry point into an extensive collection of data which is difficult to explore manually. It is possible to extend this interface to accommodate an image similarity feature. This could be useful for identifying a particular artist’s work in a broader collection.

Image classification 

While image search helps us find images, it doesn’t help us as much if we want to describe all the images in our collection. For this, we’ll need a slightly different type of machine learning task – image classification. An image classification model will put our images into categories drawn from a list of possible labels. 

We can find image classification models on the Hugging Face Hub. The “Image Classification Model Tester” tab in the demo Gradio application allows us to test most of the 3,000+ image classification models hosted on the Hub against our dataset.

This can give us a sense of a few different things:

  • How well do the labels for a model match our data?A model for classifying dog breeds probably won’t help us much!
  • It gives us a quick way of inspecting possible errors a model might make with our data. 
  • It prompts us to think about what categories might make sense for our images.
A screenshot of the image classification tab in the Gradio app which shows a bar chart with the most frequently predicted labels for images assigned by a computer vision model.

A screenshot of the image classification tab in the Gradio app which shows a bar chart with the most frequently predicted labels for images assigned by a computer vision model.

We may find a model that already does a good job working with our dataset – if we don’t, we may have to look at training a model.

Training your own computer vision model

The final tab of our Gradio demo allows you to export the image dataset in a format that can be loaded by Label Studio, an open-source tool for annotating data in preparation for machine learning tasks. In Label Studio, we can define labels we would like to apply to our dataset. For example, we might decide we’re interested in pulling out particular types of images from this collection. We can use Label Studio to create an annotated version of our dataset with these labels. This requires us to assign labels to images in our dataset with the correct labels. Although this process can take some time, it can be a useful way of further exploring a dataset and making sure your labels make sense. 

With a labeled dataset, we need some way of training a model. For this, we can use AutoTrain. This tool allows you to train machine learning models without writing any code. Using this approach supports creation of a model trained on our dataset which uses the labels we are interested in. It’s beyond the scope of this post to cover all AutoTrain features, but this post provides a useful overview of how it works.

Next Steps

As mentioned in the introduction, you can explore the ARCH Image Dataset Explorer Demo yourself. If you know a bit of Python, you could also duplicate the Space and adapt or change the current functionality it supports for exploring the dataset.

Internet Archive and Hugging Face plan to organize a hands-on hackathon later this year focused on using open source machine learning tools from the Hugging Face ecosystem to work with web archives. The event will include building interfaces for web archive datasets, collaborative annotation, and training machine learning models. Please let us know if you are interested in participating by filling out this form.

Working to Advance Library Support for Web Archive Research 

This Spring, the Internet Archive hosted two in-person workshops aimed at helping to advance library support for web archive research: Digital Scholarship & the Web and Art Resources on the Web. These one-day events were held at the Association of College & Research Libraries (ACRL) conference in Pittsburgh and the Art Libraries Society of North America (ARLIS) conference in Mexico City. The workshops brought together librarians, archivists, program officers, graduate students, and disciplinary researchers for full days of learning, discussion, and hands-on experience with web archive creation and computational analysis. The workshops were developed in collaboration with the New York Art Resources Consortium (NYARC) – and are part of an ongoing series of workshops hosted by the Internet Archive through Summer 2023.

Internet Archive Deputy Director of Archiving & Data Services Thomas Padilla discussing the potential of web archives as primary sources for computational research at Art Resources on the Web in Mexico City.

Designed in direct response to library community interest in supporting additional uses of web archive collections, the workshops had the following objectives: introduce participants to web archives as primary sources in context of computational research questions, develop familiarity with research use cases that make use of web archives; and provide an opportunity to acquire hands-on experience creating web archive collections and computationally analyzing them using ARCH (Archives Research Compute Hub) – a new service set to publicly launch June 2023.

Internet Archive Community Programs Manager Lori Donovan walking workshop participants through a demonstration of Palladio using a dataset generated with ARCH at Digital Scholarship & the Web In Pittsburgh, PA.

In support of those objectives, Internet Archive staff walked participants through web archiving workflows, introduced a diverse set of web archiving tools and technologies, and offered hands-on experience building web archives. Participants were then introduced to Archives Research Compute Hub (ARCH). ARCH supports computational research with web archive collections at scale – e.g., text and data mining, data science, digital scholarship, machine learning, and more. ARCH does this by streamlining generation and access to more than a dozen research ready web archive datasets, in-browser visualization, dataset analysis, and open dataset publication. Participants further explored data generated with ARCH in PalladioVoyant, and RAWGraphs.

Network visualization of the Occupy Web Archive collection, created using Palladio based on a Domain Graph Dataset generated by ARCH.

Gallery visualization of the CARTA Art Galleries collection, created using Palladio based on an Image Graph Dataset generated by ARCH.

At the close of the workshops, participants were eager to discuss web archive research ethics, research use cases, and a diverse set of approaches to scaling library support for researchers interested in working with web archive collections – truly vibrant discussions – and perhaps the beginnings of a community of interest!  We plan to host future workshops focused on computational research with web archives – please keep an eye on our Event Calendar.

Launching Legal Literacies for Text Data Mining – Cross Border (LLTDM-X)

We are excited to announce that the National Endowment for the Humanities (NEH) has awarded nearly $50,000 through its Digital Humanities Advancement Grant program to UC Berkeley Library and Internet Archive to study legal and ethical issues in cross-border text data mining research. NEH funding for the project, entitled Legal Literacies for Text Data Mining – Cross Border (LLTDM-X), will support research and analysis that addresses law and policy issues faced by U.S. digital humanities practitioners whose text data mining research and practice intersects with foreign-held or licensed content, or involves international research collaborations. LLTDM-X builds upon Building Legal Literacies for Text Data Mining Institute (Building LLTDM), previously funded by NEH. UC Berkeley Library directed Building LLTDM, bringing together expert faculty from across the country to train 32 digital humanities researchers on how to navigate law, policy, ethics, and risk within text data mining projects (results and impacts are summarized in the white paper here.) 

Why is LLTDM-X needed?

Text data mining, or TDM, is an increasingly essential and widespread research approach. TDM relies on automated techniques and algorithms to extract revelatory information from large sets of unstructured or thinly-structured digital content. These methodologies allow scholars to identify and analyze critical social, scientific, and literary patterns, trends, and relationships across volumes of data that would otherwise be impossible to sift through. While TDM methodologies offer great potential, they also present scholars with nettlesome law and policy challenges that can prevent them from understanding how to move forward with their research. Building LLTDM trained TDM researchers and professionals on essential principles of licensing, privacy law, as well as ethics and other legal literacies —thereby helping them move forward with impactful digital humanities research. Further, digital humanities research in particular is marked by collaboration across institutions and geographical boundaries. Yet, U.S. practitioners encounter increasingly complex cross-border problems and must accordingly consider how they work with internationally-held materials and international collaborators.

How will LLTDM-X help? 

Our long-term goal is to design instructional materials and institutes to support digital humanities TDM scholars facing cross-border issues. Through a series of virtual roundtable discussions, and accompanying legal research and analyses, LLTDM-X will surface these cross-border issues and begin to distill preliminary guidance to help scholars in navigating them. After the roundtables, we will work with the law and ethics experts to create instructive case studies that reflect the types of cross-border TDM issues practitioners encountered. Case studies, guidance, and recommendations will be widely-disseminated via an open access report to be published at the completion of the project. And most importantly, these resources will be used to inform our future educational offerings.

The LLTDM-X team is eager to get started. The project is co-directed by Thomas Padilla, Deputy Director, Archiving and Data Services at Internet Archive and Rachael Samberg, who leads UC Berkeley Library’s Office of Scholarly Communication Services. Stacy Reardon, Literatures and Digital Humanities Librarian, and Timothy Vollmer, Scholarly Communication and Copyright Librarian, both at UC Berkeley Library, round out the team.

We would like to thank NEH’s Office of Digital Humanities again for funding this important work. The full press release is available at UC Berkeley Library’s website. We invite you to contact us with any questions.