Adding pronouns, institutions, and abstracts

Author: escowles

_data/speakers.yml

@@ -16,4 +16,4 @@ speakers:
 speaker-text: Owen King, Kyeongmin Rim, Marc Verhagen
 title: "Cataloging Video via Computer Vision"
 ---
-
+Imagine you are tasked with cataloging a large stack of books.  Suppose each book has no cover, no table of contents, and no index.  And suppose copy cataloging is not an option.  Catalogers working on large collections of old, digitized videos face a similar challenge.  A video series might have been significant enough to accession, although the content of particular videos remains obscure. Watching all the videos is impractical, and scrubbing through videos is imprecise and error-prone.  Our solution has been to provide catalogers with a visual index of each video.  The visual indexes are based on a computer vision model we trained to classify scene types, especially scenes with text, from broadcast television.  This presentation explains our development and deployment of a system for creating visual indexes, including (1) our need for specialized training data, (2) our custom tooling for labeling images at maximum speed, (3) training and evaluating our computer vision model, (4) the multimedia interchange format (MMIF) for storing media annotations, (5) our methods for creating interval-based annotations from point-wise image classifications, (6) our design of visual indexes responsive to the needs of catalogers, and (7) integration of visual indexes in our metadata workflows.

_posts/2025-03-10-cataloging-video-via-computer-vision.md

@@ -14,4 +14,4 @@ speakers:
 speaker-text: Beck Davis, , 
 title: "Enhancing MARC Language Metadata with Ruby and Indigenous Tenacity"
 ---
-
+There are many Indigenous and minority languages that are grouped in broader MARC language codes, limiting the ability to find materials in a specific language. I was inspired by the University of Hawaii’s Indigenous Pacific Languages cataloging project, which highlighted the limitations of the MARC Language Codes. I undertook this project to improve language-related discoverability by parsing free text language metadata from MARC field 546 to create new 041 fields with matching ISO 639-3 language codes. The goal was improving the discoverability of Indigenous and minority language materials by enhancing language metadata accuracy. ISO 639-3 language codes can supplement MARC Language Codes for more accurate language representation. In the process, I identified many candidate materials in the library’s collection. This talk will discuss the importance of supplementing with ISO 639-3 language codes, as well as the challenges I encountered as I developed software tools to make that easier for catalogers. 

_posts/2025-03-10-enhancing-marc-language-metadata-with-ruby-and-indigenous-tenacity.md

@@ -13,6 +13,6 @@ speakers:
 - lauren-gallina
 - john-dewees
 speaker-text: Lauren Gallina, John Dewees, 
-title: "Getting Everyone on the Same Page: Building a Shared Governance Structure for Digital Projects"
+title: "Getting Everyone on the Same Page: Building a Shared Governance Structure for Digital Projects "
 ---
-
+Coordinating the work of multiple people on projects can be a thorny problem. This can be especially difficult when folks work in different departments and on different levels of the org chart. The University of Rochester has been working on a governance structure to ensure that the handling of digital projects can be done consistently, that the management of these projects does not depend on a particular person being in a role, and to foster clear communication and greater trust. This presentation will include a discussion of how this iterative process came about, how the framework developed for digital projects management came to be, what artifacts were necessary to create along the way, and how we are going to guarantee that this model will keep us on track and help accomplish our goals well into the future. 

_posts/2025-03-10-getting-everyone-on-the-same-page-building-a-shared-governance-structure-for-digital-projects.md

@@ -14,4 +14,4 @@ speakers:
 speaker-text: Sarah Hensler, , 
 title: "If You Give a Girl a Website (She'll ask for metadata searching)"
 ---
-
+When non-library platforms want to integrate online archival materials into their project, they often turn to pre-made web content management systems to lessen the need for advanced coding and metadata skills. This was the case for Windows to the World (W2W), a digital humanities curriculum platform that uses material from several different archival repositories. The W2W website, which was designed by university WCMS developers, has very limited metadata functionalities for featured museum items and no metadata for the curricula. For a platform that aims to make creative educational connections between otherwise disparate primary sources, this has become a crushing problem. However, a new grant- and a young MLS student learning to code- offer the chance to change that. This talk showcases one MLS student’s efforts to transition a non-functional WCMS into a platform where metadata drives educational engagement. Using a combination of library science pedagogy, inter-institutional collaborations, and some hot-off-the-press skills in GitHub CollectionBuilder and HTML, this grad student explores how librarians can reframe educators’ understanding of archival materials when they prioritize the metadata around each item. 

_posts/2025-03-10-if-you-give-a-girl-a-website-she-ll-ask-for-metadata-searching.md

@@ -14,4 +14,4 @@ speakers:
 speaker-text: Bess Sadler, , 
 title: "Infosec for a world at war"
 ---
-
+What does it mean to host digitized library collections that state actors would like to destroy? Or to have custody of data and documents that are evidence in a political battle? What responsibilities do we have to Justice, when we might hold evidence relevant to land back lawsuits, which powerful interests would like to destroy? This talk will discuss how to apply agile and DevOps principles to the question of information security for digital libraries, and provide some examples about why it matters.

_posts/2025-03-10-infosec-for-a-world-at-war.md

@@ -14,4 +14,10 @@ speakers:
 speaker-text: Monique Rio, , 
 title: "LC Subject Heading Remapping: Making Good on Good Intentions"
 ---
+Library of Congress subject headings don't always reflect the values of the institutions using them, and they change at the speed of Congress. Over the years there has been a desire for institutions to choose for themselves what words to use for LC terms, and there have been a variety of ways to do this.
 
+The University of Michigan used the method of creating local authority records to replace the terms. This worked well for our records, but because we include records that aren't ours--from the Alma community zone and from HathiTrust--these changes weren't applied to all of the records in our catalog. This meant that for users using our discovery system, instead of only seeing our preferred terms, they had a mix of our preferred term AND the current LC term. On top of this, University of Michigan has an LC Subject Browse application, and all of our preferred terms weren't included--because they aren't technically LC terms anymore--effectively downgrading their usefulness. 
+
+With inspiration from a presentation by Triangle Research Libraries Network for the ALA in 2021, we came up with a plan to thoroughly fix this problem. Right before the Fall 2024 term, the solution went into use. We went into the project thinking it would be a simple "replace this phrase with that phrase" problem, and it turned out to be quite a bit more complicated than that.
+
+This presentation will discuss why this is a hard problem, how we fixed the problem, and what institutional structures enabled us to actually get this work done.

_posts/2025-03-10-lc-subject-heading-remapping-making-good-on-good-intentions.md

@@ -14,4 +14,4 @@ speakers:
 speaker-text: Katherine "Kate" Deibel, , 
 title: "Making the Development Process Inclusive of Disabilities"
 ---
-
+With accessibility, we often think about the finished technology. Disabled people, however, are not just users. Many developers, managers, testers, etc. have disabilities that our work practices may be problematic but addressable. This talk will highlight several common practices and tools used in development (e.g., Slack, agile) and how to make them more inclusive for sensory disabilities and neurodivergence.

_posts/2025-03-10-making-the-development-process-inclusive-of-disabilities.md

@@ -14,4 +14,4 @@ speakers:
 speaker-text: Charlotte Kostelic, , 
 title: "MARC record validation with Python and Pydantic"
 ---
-
+Pydantic is a widely used data validation library for Python favored for its speed and extensibility. With Pydantic users can define schema or models against which to validate data. Users receive detailed reports of any validation errors after passing data into or validating a data object against a model. In my presentation I will outline how I have used Pydantic to validate vendor-supplied MARC records to improve their quality. As NYPL has expanded the use of shelf-ready services for some research materials, we have increased the points at which we can introduce inconsistencies into our catalog. In this presentation I will describe how to define models to validate embedded order data (EOD), highlight how to customize models to account for slight differences between required fields in vendor records, and outline how to create one’s own Pydantic models. In addition to describing my pilot project to validate vendor-supplied data for shelf-ready MARC records, I will also outline how one can use Pydantic to validate the structure of MARC records. 

_posts/2025-03-10-marc-record-validation-with-python-and-pydantic.md

@@ -15,4 +15,4 @@ speakers:
 speaker-text: Brandon Patterson, Matthew Elliott, 
 title: "Preserving History Through Advanced 3D Scanning: A Case Study on Gaussian Splat Photogrammetry"
 ---
-
+In the realm of historical preservation, innovative technologies play a crucial role in capturing and safeguarding the essence of culturally significant objects and structures. This presentation highlights a cutting-edge 3D scanning project undertaken by our library team, wherein we took over 3,600 photos of multiple rooms of a historically significant building using a novel photogrammetry technique. At the core of our method lies the Gaussian Splat approach, which facilitates real-time radiance field rendering, resulting in highly detailed and realistic 3D models. Our project not only emphasizes the practical applications of this advanced technology in historical documentation but also showcases the challenges and successes encountered during the process. Through this initiative, we aim to inspire further exploration of photogrammetry in cultural heritage conservation and encourage interdisciplinary collaborations between libraries, coding experts, and preservationists.