The Cyberne*cs Thought Collec*ve Project

Transcription

1 The Cyberne*cs Thought Collec*ve Project Using Computa*onal Methods to Reveal Intellectual Context in Archival Material Bethany G. Anderson, Christopher J. Prom, Kevin Hamilton, James A. Hutchinson, Mark Sammons, and Alex Dolski University of Illinois at Urbana- Champaign IEEE Big Data 2017 Computa*onal Archival Science Workshop, December 13, 2017

2 Special Thanks to Anirudh Chandrashekhar (Undergraduate Student Programmer, University of Illinois at Urbana- Champaign)

3 The CyberneLcs Thought CollecLve: Project Overview Pilot grant from the Na*onal Endowment for the Humani*es (Humani*es Collec*ons and Reference Resources program), May April 2018 Collabora*ve project between four ins*tu*ons that hold the research archives of founding members of cyberne*cs The Heinz von Foerster Papers at the University of Illinois The W. Ross Ashby Papers at the British Library (Ashby right) The Norbert Wiener Papers at MIT The Warren S. McCulloch Papers at the American Philosophical Society

4 Project website: hvps://archives.library.illinois.edu/thought- collec*ve/

5 Proposed Work Digitize Archival Materials Metadata Creation and Remediation Preservation and basic access Testing annotation and entity extraction tools This work will result in a prototype thought colleclve portal enable users to explore the cybernelcs movement in ways that would be impossible with tradilonal systems, via new interfaces that use machine learning techniques to reveal previously hidden relalonships between people, topics, and localons.

6 CyberneLcs: A Postwar ScienLfic Movement Macy Conferences on Circular Causal and Feedback Mechanisms in Biological and Social Systems - later Macy Conferences on Cyberne*cs a\er Wiener s publica*on ( ) Norbert Wiener s Cyberne'cs: Or Control and Communica'on in the Animal and Machine (1948) Cover from the Transactions of the 10th Macy Conference on Cybernetics, 1953 (edited by Heinz von Foerster, Margaret Mead, and Hans Lukas Teuber). From the Heinz von Foerster Papers, courtesy of the University of Illinois Archives.

7 Origins of the CyberneLcs Thought CollecLve: The Macy Conferences

8 A science with many definilons As the science of communica*on and control, or steersmanship, in organisms and machines (N. Wiener, 1948) As the study of form and pavern (G. Bateson, 1972) As the study of behavior (W. R. Ashby, 1956) Images from the Heinz von Foerster Papers, courtesy of the University of Illinois Archives

9 Applying ComputaLonal Methods to Archives

10 Employing Natural Language Processing and Machine Learning Technologies Archives and digital humani*es projects are increasingly employing natural language processing (NLP), named en*ty recogni*on (NER), and machine learning technologies - e.g., Carnegie Mellon University Archives, epadd, BitCurator NLP, Woodseer, ArchExtract Artificial Intelligence : Image from the Heinz von Foerster Papers, courtesy of the University of Illinois Archives The Cyberne*cs Thought Collec*ve (CTC) team is applying NLP and NER in addi*on to neural network (NN) technologies to digi*zed material from the personal archives of von Foerster, McCulloch, and Wiener, and the scien*fic journals of Ashby

11 Defining CyberneLc EnLLes Word cloud of Cybernetics of Cybernetics from Voyant List of cybernetics terms created from Voyant list with feedback from advisory board

12 Work To Date: Exploring and TesLng SoYware

13 Work To Date: Frequency and N- Grams

14 Work To Date: ClassificaLon and Clustering

15 Short- term Technical Work: Metadata

16 Short- term Technical Work: Digital Repos/Library

17 Short- term Technical Work: Portal UI?

18 Short- term Technical Work: RepresenLng Certainty

19 Outstanding issues: RepresenLng ML results

20 Outstanding issue: Metadata VisualizaLons

21 Longer- term Project Plans Using machine learning to transcribe handwriven texts and annota*ons (backburner) Define the boundaries of the thought collec*ve Submit proposal for NEH Implementa*on grant (Humani*es Collec*ons and Reference Resources) Include other cyberne*cs collec*ons Create model for ins*tu*ons to contribute collec*ons Apply lessons to other corpora

22 Project Takeaways and Discussion Items Aggrega*ons and Tradi*onal Archival Work s*ll maver Grouping by provenance/hierarchy Some metadata (from, to) applied from folder hierarchy Dates hand recorded Concepts machine extracted Methods for digital a model for textual documents? Which approach generates more authen*c connec*ons in the thought collec*ve - supervised or unsupervised? Can we ensure machine recognizes nuances and concepts which were called different things at different *mes?

23 Thank you! Bethany Anderson, Christopher Prom,