Presentations
Permanent URI for this collection
Browse
Recent Submissions
Publication Sociotechnical Automation Science: A Case Study in Developing and Augmenting an Ensemble Neural Network with Multiple LLMs for Subject Cataloging at the Penn Libraries(2024-06-26) Hahn, JimThe sociotechnical aspects of automation play a crucial role in the development of machine learning systems. Through deep collaboration with cataloging professionals at the Penn Libraries, we have created a set of subject indexing algorithms that are ensembled into a neural network. Librarians have evaluated multiple rounds of the algorithm outputs. By identifying the failure points in the neural network-based subject assignment process, we incorporated LLM tasks such as evaluating search result relevance, summarizing search results, and assessing topical assignments of synthetic summaries. Implementing LLM tasks draws on the linguistic strengths of LLMs, rather than world knowledge. The data processing is integrated into an Apache Airflow pipeline, allowing librarians to input an Excel file, which begins the workflow for generating candidate subject descriptions. These machine learning outputs are poised for a pilot test in production systems this summer.Publication Online Optimization of Soft Manipulator Mechanics via Hierarchical Control(2024) Misra, Shivangi; Sung, CynthiaActively tuning mechanical properties in soft robots is now feasible due to advancements in soft actuation technologies. In soft manipulators, these novel actuators can be distributed over the robot body to allow greater control over its large number of degrees of freedom and to stabilize local deformations against a range of disturbances. In this paper, we present a hierarchical policy for stiffness control for such a class of soft manipulators. The stiffness changes induce desired deformations in each segment, thereby influencing the manipulator’s end-effector position. The algorithm can be run as an online controller to influence the manipulator’s stable states – as we demonstrate in simulation – or offline as a design algorithm to optimize stiffness distributions – as we showcase in a hardware demonstration. Our proposed hierarchical control scheme is agnostic to the stiffness actuation method and can extend to other soft manipulators with nonuniform stiffness distributions.Publication Audiovisual Data Curation Primer Presentation(2023-12-14) Phegley, LaurenThis presentation was given as part of the Data Curation Network's Primer Webinar held on 2023-12-14. The authors presented the highlights of our Audiovisual Data Curation Primer, which is a peer-reviewed concise resource designed to provide support for data curators in learning about audiovisual files. The full primer is openly avaliable at https://github.com/DataCurationNetwork/data-primers/blob/master/Audiovisual%20Data%20Curation%20Primer/AV-data-curation-primer.md.Publication BIBFRAME instance mining: Toward authoritative publisher entities using association rules(2020-11-25)With the transition of a shared catalog to BIBFRAME linked data, there is now a pressing need for identifying the canonical Instance for clustering in BIBFRAME. A fundamental component of Instance identification is by way of authoritative publisher entities. Previous work in this area by OCLC research (Connaway & Dickey, 2011) proposed a data mining approach for developing an experimental Publisher Name Authority File (PNAF). The OCLC research was able to create profiles for "high-incidence" publishers after data mining and clustering of publishers. As a component of PNAF, Connaway & Dickney were able to provide detailed subject analysis of publishers. This presentation will detail a case study of machine learning methods over a corpus of subjects, main entries, and added entries, as antecedents into association rules to derive consequent publisher entities. The departure point for the present research into identification of authoritative publisher entities is to focus on clustering, reconciliation and re-use of ISBN and subfield b of MARC 260 along with the subjects (650 - Subject Added Entry), main entries (1XX - Main Entries) and added entries (710 - Added Entry-Corporate Name) as signals to inform a training corpus into association rule mining, among other machine learning algorithms, libraries, and methods.Publication BF Interlingua: Interoperability among BIBFRAME linked data vocabularies(2023-01-19)Presentation exploring an interchange process among BIBFRAME linked data vocabularies.